Datasets:

IAMJB

/

EMNLP2023-papers

like 0

https://aclanthology.org/2023.emnlp-main.401.bib

https://aclanthology.org/2023.emnlp-main.401/

@inproceedings{weissweiler-etal-2023-counting, title = "Counting the Bugs in {C}hat{GPT}{'}s Wugs: A Multilingual Investigation into the Morphological Capabilities of a Large Language Model", author = "Weissweiler, Leonie and Hofmann, Valentin and Kantharuban, Anjali and Cai, Anna and Dutt, Ritam and Hengle, Amey and Kabra, Anubha and Kulkarni, Atharva and Vijayakumar, Abhishek and Yu, Haofei and Schuetze, Hinrich and Oflazer, Kemal and Mortensen, David", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.401", doi = "10.18653/v1/2023.emnlp-main.401", pages = "6508--6524", abstract = "Large language models (LLMs) have recently reached an impressive level of linguistic capability, prompting comparisons with human language skills. However, there have been relatively few systematic inquiries into the linguistic capabilities of the latest generation of LLMs, and those studies that do exist (i) ignore the remarkable ability of humans to generalize, (ii) focus only on English, and (iii) investigate syntax or semantics and overlook other capabilities that lie at the heart of human language, like morphology. Here, we close these gaps by conducting the first rigorous analysis of the morphological capabilities of ChatGPT in four typologically varied languages (specifically, English, German, Tamil, and Turkish). We apply a version of Berko{'}s (1958) wug test to ChatGPT, using novel, uncontaminated datasets for the four examined languages. We find that ChatGPT massively underperforms purpose-built systems, particularly in English. Overall, our results{---}through the lens of morphology{---}cast a new light on the linguistic capabilities of ChatGPT, suggesting that claims of human-like language skills are premature and misleading.", }

Large language models (LLMs) have recently reached an impressive level of linguistic capability, prompting comparisons with human language skills. However, there have been relatively few systematic inquiries into the linguistic capabilities of the latest generation of LLMs, and those studies that do exist (i) ignore the remarkable ability of humans to generalize, (ii) focus only on English, and (iii) investigate syntax or semantics and overlook other capabilities that lie at the heart of human language, like morphology. Here, we close these gaps by conducting the first rigorous analysis of the morphological capabilities of ChatGPT in four typologically varied languages (specifically, English, German, Tamil, and Turkish). We apply a version of Berko{'}s (1958) wug test to ChatGPT, using novel, uncontaminated datasets for the four examined languages. We find that ChatGPT massively underperforms purpose-built systems, particularly in English. Overall, our results{---}through the lens of morphology{---}cast a new light on the linguistic capabilities of ChatGPT, suggesting that claims of human-like language skills are premature and misleading.

Counting the Bugs in ChatGPT's Wugs: A Multilingual Investigation into the Morphological Capabilities of a Large Language Model

emnlp-main.401

2310.15113

Poster

https://aclanthology.org/2023.emnlp-main.402.bib

https://aclanthology.org/2023.emnlp-main.402/

@inproceedings{long-etal-2023-adapt, title = "Adapt in Contexts: Retrieval-Augmented Domain Adaptation via In-Context Learning", author = "Long, Quanyu and Wang, Wenya and Pan, Sinno", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.402", doi = "10.18653/v1/2023.emnlp-main.402", pages = "6525--6542", abstract = "Large language models (LLMs) have showcased their capability with few-shot inference known as in-context learning. However, in-domain demonstrations are not always readily available in real scenarios, leading to cross-domain in-context learning. Besides, LLMs are still facing challenges in long-tail knowledge in unseen and unfamiliar domains. The above limitations demonstrate the necessity of Unsupervised Domain Adaptation (UDA). In this paper, we study the UDA problem under an in-context learning setting to adapt language models from the source domain to the target domain without any target labels. The core idea is to retrieve a subset of cross-domain elements that are the most similar to the query, and elicit language model to adapt in an in-context manner by learning both target domain distribution and the discriminative task signal simultaneously with the augmented cross-domain in-context examples. We devise different prompting and training strategies, accounting for different LM architectures to learn the target distribution via language modeling. With extensive experiments on Sentiment Analysis (SA) and Named Entity Recognition (NER) tasks, we thoroughly study the effectiveness of ICL for domain transfer and demonstrate significant improvements over baseline models.", }

Large language models (LLMs) have showcased their capability with few-shot inference known as in-context learning. However, in-domain demonstrations are not always readily available in real scenarios, leading to cross-domain in-context learning. Besides, LLMs are still facing challenges in long-tail knowledge in unseen and unfamiliar domains. The above limitations demonstrate the necessity of Unsupervised Domain Adaptation (UDA). In this paper, we study the UDA problem under an in-context learning setting to adapt language models from the source domain to the target domain without any target labels. The core idea is to retrieve a subset of cross-domain elements that are the most similar to the query, and elicit language model to adapt in an in-context manner by learning both target domain distribution and the discriminative task signal simultaneously with the augmented cross-domain in-context examples. We devise different prompting and training strategies, accounting for different LM architectures to learn the target distribution via language modeling. With extensive experiments on Sentiment Analysis (SA) and Named Entity Recognition (NER) tasks, we thoroughly study the effectiveness of ICL for domain transfer and demonstrate significant improvements over baseline models.

Adapt in Contexts: Retrieval-Augmented Domain Adaptation via In-Context Learning

emnlp-main.402

2311.11551

Poster

https://aclanthology.org/2023.emnlp-main.403.bib

https://aclanthology.org/2023.emnlp-main.403/

@inproceedings{brown-etal-2023-understanding, title = "Understanding the Inner-workings of Language Models Through Representation Dissimilarity", author = "Brown, Davis and Godfrey, Charles and Konz, Nicholas and Tu, Jonathan and Kvinge, Henry", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.403", doi = "10.18653/v1/2023.emnlp-main.403", pages = "6543--6558", abstract = "As language models are applied to an increasing number of real-world applications, understanding their inner workings has become an important issue in model trust, interpretability, and transparency. In this work we show that representation dissimilarity measures, which are functions that measure the extent to which two model{'}s internal representations differ, can be a valuable tool for gaining insight into the mechanics of language models. Among our insights are: (i) an apparent asymmetry in the internal representations of model using SoLU and GeLU activation functions, (ii) evidence that dissimilarity measures can identify and locate generalization properties of models that are invisible via in-distribution test set performance, and (iii) new evaluations of how language model features vary as width and depth are increased. Our results suggest that dissimilarity measures are a promising set of tools for shedding light on the inner workings of language models.", }

As language models are applied to an increasing number of real-world applications, understanding their inner workings has become an important issue in model trust, interpretability, and transparency. In this work we show that representation dissimilarity measures, which are functions that measure the extent to which two model{'}s internal representations differ, can be a valuable tool for gaining insight into the mechanics of language models. Among our insights are: (i) an apparent asymmetry in the internal representations of model using SoLU and GeLU activation functions, (ii) evidence that dissimilarity measures can identify and locate generalization properties of models that are invisible via in-distribution test set performance, and (iii) new evaluations of how language model features vary as width and depth are increased. Our results suggest that dissimilarity measures are a promising set of tools for shedding light on the inner workings of language models.

Understanding the Inner-workings of Language Models Through Representation Dissimilarity

emnlp-main.403

2310.14993

Poster

https://aclanthology.org/2023.emnlp-main.404.bib

https://aclanthology.org/2023.emnlp-main.404/

@inproceedings{lu-etal-2023-efficient, title = "Efficient Classification of Long Documents via State-Space Models", author = "Lu, Peng and Wang, Suyuchen and Rezagholizadeh, Mehdi and Liu, Bang and Kobyzev, Ivan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.404", doi = "10.18653/v1/2023.emnlp-main.404", pages = "6559--6565", abstract = "Transformer-based models have achieved state-of-the-art performance on numerous NLP applications. However, long documents which are prevalent in real-world scenarios cannot be efficiently processed by transformers with the vanilla self-attention module due to their quadratic computation complexity and limited length extrapolation ability. Instead of tackling the computation difficulty for self-attention with sparse or hierarchical structures, in this paper, we investigate the use of State-Space Models (SSMs) for long document classification tasks. We conducted extensive experiments on six long document classification datasets, including binary, multi-class, and multi-label classification, comparing SSMs (with and without pre-training) to self-attention-based models. We also introduce the SSM-pooler model and demonstrate that it achieves comparable performance while being on average 36{\%} more efficient. Additionally our method exhibits higher robustness to the input noise even in the extreme scenario of 40{\%}.", }

Transformer-based models have achieved state-of-the-art performance on numerous NLP applications. However, long documents which are prevalent in real-world scenarios cannot be efficiently processed by transformers with the vanilla self-attention module due to their quadratic computation complexity and limited length extrapolation ability. Instead of tackling the computation difficulty for self-attention with sparse or hierarchical structures, in this paper, we investigate the use of State-Space Models (SSMs) for long document classification tasks. We conducted extensive experiments on six long document classification datasets, including binary, multi-class, and multi-label classification, comparing SSMs (with and without pre-training) to self-attention-based models. We also introduce the SSM-pooler model and demonstrate that it achieves comparable performance while being on average 36{\%} more efficient. Additionally our method exhibits higher robustness to the input noise even in the extreme scenario of 40{\%}.

Efficient Classification of Long Documents via State-Space Models

emnlp-main.404

Poster

https://aclanthology.org/2023.emnlp-main.405.bib

https://aclanthology.org/2023.emnlp-main.405/

@inproceedings{shi-etal-2023-dual, title = "Dual-Feedback Knowledge Retrieval for Task-Oriented Dialogue Systems", author = "Shi, Tianyuan and Li, Liangzhi and Lin, Zijian and Yang, Tao and Quan, Xiaojun and Wang, Qifan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.405", doi = "10.18653/v1/2023.emnlp-main.405", pages = "6566--6580", abstract = "Efficient knowledge retrieval plays a pivotal role in ensuring the success of end-to-end task-oriented dialogue systems by facilitating the selection of relevant information necessary to fulfill user requests. However, current approaches generally integrate knowledge retrieval and response generation, which poses scalability challenges when dealing with extensive knowledge bases. Taking inspiration from open-domain question answering, we propose a retriever-generator architecture that harnesses a retriever to retrieve pertinent knowledge and a generator to generate system responses. Due to the lack of retriever training labels, we propose relying on feedback from the generator as pseudo-labels to train the retriever. To achieve this, we introduce a dual-feedback mechanism that generates both positive and negative feedback based on the output of the generator. Our method demonstrates superior performance in task-oriented dialogue tasks, as evidenced by experimental results on three benchmark datasets.", }

Efficient knowledge retrieval plays a pivotal role in ensuring the success of end-to-end task-oriented dialogue systems by facilitating the selection of relevant information necessary to fulfill user requests. However, current approaches generally integrate knowledge retrieval and response generation, which poses scalability challenges when dealing with extensive knowledge bases. Taking inspiration from open-domain question answering, we propose a retriever-generator architecture that harnesses a retriever to retrieve pertinent knowledge and a generator to generate system responses. Due to the lack of retriever training labels, we propose relying on feedback from the generator as pseudo-labels to train the retriever. To achieve this, we introduce a dual-feedback mechanism that generates both positive and negative feedback based on the output of the generator. Our method demonstrates superior performance in task-oriented dialogue tasks, as evidenced by experimental results on three benchmark datasets.

Dual-Feedback Knowledge Retrieval for Task-Oriented Dialogue Systems

emnlp-main.405

2310.14528

Poster

https://aclanthology.org/2023.emnlp-main.406.bib

https://aclanthology.org/2023.emnlp-main.406/

@inproceedings{boisson-etal-2023-construction, title = "Construction Artifacts in Metaphor Identification Datasets", author = "Boisson, Joanne and Espinosa-Anke, Luis and Camacho-Collados, Jose", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.406", doi = "10.18653/v1/2023.emnlp-main.406", pages = "6581--6590", abstract = "Metaphor identification aims at understanding whether a given expression is used figuratively in context. However, in this paper we show how existing metaphor identification datasets can be gamed by fully ignoring the potential metaphorical expression or the context in which it occurs. We test this hypothesis in a variety of datasets and settings, and show that metaphor identification systems based on language models without complete information can be competitive with those using the full context. This is due to the construction procedures to build such datasets, which introduce unwanted biases for positive and negative classes. Finally, we test the same hypothesis on datasets that are carefully sampled from natural corpora and where this bias is not present, making these datasets more challenging and reliable.", }

Metaphor identification aims at understanding whether a given expression is used figuratively in context. However, in this paper we show how existing metaphor identification datasets can be gamed by fully ignoring the potential metaphorical expression or the context in which it occurs. We test this hypothesis in a variety of datasets and settings, and show that metaphor identification systems based on language models without complete information can be competitive with those using the full context. This is due to the construction procedures to build such datasets, which introduce unwanted biases for positive and negative classes. Finally, we test the same hypothesis on datasets that are carefully sampled from natural corpora and where this bias is not present, making these datasets more challenging and reliable.

Construction Artifacts in Metaphor Identification Datasets

emnlp-main.406

2311.00790

Poster

https://aclanthology.org/2023.emnlp-main.407.bib

https://aclanthology.org/2023.emnlp-main.407/

@inproceedings{nathani-etal-2023-maf, title = "{MAF}: Multi-Aspect Feedback for Improving Reasoning in Large Language Models", author = "Nathani, Deepak and Wang, David and Pan, Liangming and Wang, William", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.407", doi = "10.18653/v1/2023.emnlp-main.407", pages = "6591--6616", abstract = "Language Models (LMs) have shown impressive performance in various natural language tasks. However, when it comes to natural language reasoning, LMs still face challenges such as hallucination, generating incorrect intermediate reasoning steps, and making mathematical errors. Recent research has focused on enhancing LMs through *self-improvement* using feedback. Nevertheless, existing approaches relying on a single generic feedback source fail to address the diverse error types found in LM-generated reasoning chains. In this work, we propose **Multi-Aspect Feedback**, an iterative refinement framework that integrates multiple feedback modules, including frozen LMs and external tools, each focusing on a specific error category. Our experimental results demonstrate the efficacy of our approach to addressing several errors in the LM-generated reasoning chain and thus improving the overall performance of an LM in several reasoning tasks. We see an improvement of up to 20{\%} in Mathematical Reasoning and up to 18{\%} in Logical Entailment.", }

Language Models (LMs) have shown impressive performance in various natural language tasks. However, when it comes to natural language reasoning, LMs still face challenges such as hallucination, generating incorrect intermediate reasoning steps, and making mathematical errors. Recent research has focused on enhancing LMs through *self-improvement* using feedback. Nevertheless, existing approaches relying on a single generic feedback source fail to address the diverse error types found in LM-generated reasoning chains. In this work, we propose **Multi-Aspect Feedback**, an iterative refinement framework that integrates multiple feedback modules, including frozen LMs and external tools, each focusing on a specific error category. Our experimental results demonstrate the efficacy of our approach to addressing several errors in the LM-generated reasoning chain and thus improving the overall performance of an LM in several reasoning tasks. We see an improvement of up to 20{\%} in Mathematical Reasoning and up to 18{\%} in Logical Entailment.

MAF: Multi-Aspect Feedback for Improving Reasoning in Large Language Models

emnlp-main.407

2310.12426

https://huggingface.co/papers/2310.12426

Poster

https://aclanthology.org/2023.emnlp-main.408.bib

https://aclanthology.org/2023.emnlp-main.408/

@inproceedings{shi-etal-2023-granularity, title = "Granularity Matters: Pathological Graph-driven Cross-modal Alignment for Brain {CT} Report Generation", author = "Shi, Yanzhao and Ji, Junzhong and Zhang, Xiaodan and Qu, Liangqiong and Liu, Ying", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.408", doi = "10.18653/v1/2023.emnlp-main.408", pages = "6617--6630", abstract = "The automatic Brain CT reports generation can improve the efficiency and accuracy of diagnosing cranial diseases. However, current methods are limited by 1) coarse-grained supervision: the training data in image-text format lacks detailed supervision for recognizing subtle abnormalities, and 2) coupled cross-modal alignment: visual-textual alignment may be inevitably coupled in a coarse-grained manner, resulting in tangled feature representation for report generation. In this paper, we propose a novel Pathological Graph-driven Cross-modal Alignment (PGCA) model for accurate and robust Brain CT report generation. Our approach effectively decouples the cross-modal alignment by constructing a Pathological Graph to learn fine-grained visual cues and align them with textual words. This graph comprises heterogeneous nodes representing essential pathological attributes (i.e., tissue and lesion) connected by intra- and inter-attribute edges with prior domain knowledge. Through carefully designed graph embedding and updating modules, our model refines the visual features of subtle tissues and lesions and aligns them with textual words using contrastive learning. Extensive experimental results confirm the viability of our method. We believe that our PGCA model holds the potential to greatly enhance the automatic generation of Brain CT reports and ultimately contribute to improved cranial disease diagnosis.", }

The automatic Brain CT reports generation can improve the efficiency and accuracy of diagnosing cranial diseases. However, current methods are limited by 1) coarse-grained supervision: the training data in image-text format lacks detailed supervision for recognizing subtle abnormalities, and 2) coupled cross-modal alignment: visual-textual alignment may be inevitably coupled in a coarse-grained manner, resulting in tangled feature representation for report generation. In this paper, we propose a novel Pathological Graph-driven Cross-modal Alignment (PGCA) model for accurate and robust Brain CT report generation. Our approach effectively decouples the cross-modal alignment by constructing a Pathological Graph to learn fine-grained visual cues and align them with textual words. This graph comprises heterogeneous nodes representing essential pathological attributes (i.e., tissue and lesion) connected by intra- and inter-attribute edges with prior domain knowledge. Through carefully designed graph embedding and updating modules, our model refines the visual features of subtle tissues and lesions and aligns them with textual words using contrastive learning. Extensive experimental results confirm the viability of our method. We believe that our PGCA model holds the potential to greatly enhance the automatic generation of Brain CT reports and ultimately contribute to improved cranial disease diagnosis.

Granularity Matters: Pathological Graph-driven Cross-modal Alignment for Brain CT Report Generation

emnlp-main.408

Oral

https://aclanthology.org/2023.emnlp-main.409.bib

https://aclanthology.org/2023.emnlp-main.409/

@inproceedings{wu-etal-2023-enhancing, title = "Enhancing Structured Evidence Extraction for Fact Verification", author = "Wu, Zirui and Hu, Nan and Feng, Yansong", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.409", doi = "10.18653/v1/2023.emnlp-main.409", pages = "6631--6641", abstract = "Open-domain fact verification is the task of verifying claims in natural language texts against extracted evidence. FEVEROUS is a benchmark that requires extracting and integrating both unstructured and structured evidence to verify a given claim. Previous models suffer from low recall of structured evidence extraction, i.e., table extraction and cell selection. In this paper, we propose a simple but effective method to enhance the extraction of structured evidence by leveraging the row and column semantics of tables. Our method comprises two components: (i) a coarse-grained table extraction module that selects tables based on rows and columns relevant to the claim and (ii) a fine-grained cell selection graph that combines both formats of evidence and enables multi-hop and numerical reasoning. We evaluate our method on FEVEROUS and achieve an evidence recall of 60.01{\%} on the test set, which is 6.14{\%} higher than the previous state-of-the-art performance. Our results demonstrate that our method can extract tables and select cells effectively, and provide better evidence sets for verdict prediction. Our code is released at https://github.com/WilliamZR/see-st", }

Open-domain fact verification is the task of verifying claims in natural language texts against extracted evidence. FEVEROUS is a benchmark that requires extracting and integrating both unstructured and structured evidence to verify a given claim. Previous models suffer from low recall of structured evidence extraction, i.e., table extraction and cell selection. In this paper, we propose a simple but effective method to enhance the extraction of structured evidence by leveraging the row and column semantics of tables. Our method comprises two components: (i) a coarse-grained table extraction module that selects tables based on rows and columns relevant to the claim and (ii) a fine-grained cell selection graph that combines both formats of evidence and enables multi-hop and numerical reasoning. We evaluate our method on FEVEROUS and achieve an evidence recall of 60.01{\%} on the test set, which is 6.14{\%} higher than the previous state-of-the-art performance. Our results demonstrate that our method can extract tables and select cells effectively, and provide better evidence sets for verdict prediction. Our code is released at https://github.com/WilliamZR/see-st

Enhancing Structured Evidence Extraction for Fact Verification

emnlp-main.409

Poster

https://aclanthology.org/2023.emnlp-main.410.bib

https://aclanthology.org/2023.emnlp-main.410/

@inproceedings{wu-etal-2023-rethinking-model, title = "Rethinking Model Selection and Decoding for Keyphrase Generation with Pre-trained Sequence-to-Sequence Models", author = "Wu, Di and Ahmad, Wasi and Chang, Kai-Wei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.410", doi = "10.18653/v1/2023.emnlp-main.410", pages = "6642--6658", abstract = "Keyphrase Generation (KPG) is a longstanding task in NLP with widespread applications. The advent of sequence-to-sequence (seq2seq) pre-trained language models (PLMs) has ushered in a transformative era for KPG, yielding promising performance improvements. However, many design decisions remain unexplored and are often made arbitrarily. This paper undertakes a systematic analysis of the influence of model selection and decoding strategies on PLM-based KPG. We begin by elucidating why seq2seq PLMs are apt for KPG, anchored by an attention-driven hypothesis. We then establish that conventional wisdom for selecting seq2seq PLMs lacks depth: (1) merely increasing model size or performing task-specific adaptation is not parameter-efficient; (2) although combining in-domain pre-training with task adaptation benefits KPG, it does partially hinder generalization. Regarding decoding, we demonstrate that while greedy search achieves strong F1 scores, it lags in recall compared with sampling-based methods. Based on these insights, we propose DeSel, a likelihood-based decode-select algorithm for seq2seq PLMs. DeSel improves greedy search by an average of 4.7{\%} semantic F1 across five datasets. Our collective findings pave the way for deeper future investigations into PLM-based KPG.", }

Keyphrase Generation (KPG) is a longstanding task in NLP with widespread applications. The advent of sequence-to-sequence (seq2seq) pre-trained language models (PLMs) has ushered in a transformative era for KPG, yielding promising performance improvements. However, many design decisions remain unexplored and are often made arbitrarily. This paper undertakes a systematic analysis of the influence of model selection and decoding strategies on PLM-based KPG. We begin by elucidating why seq2seq PLMs are apt for KPG, anchored by an attention-driven hypothesis. We then establish that conventional wisdom for selecting seq2seq PLMs lacks depth: (1) merely increasing model size or performing task-specific adaptation is not parameter-efficient; (2) although combining in-domain pre-training with task adaptation benefits KPG, it does partially hinder generalization. Regarding decoding, we demonstrate that while greedy search achieves strong F1 scores, it lags in recall compared with sampling-based methods. Based on these insights, we propose DeSel, a likelihood-based decode-select algorithm for seq2seq PLMs. DeSel improves greedy search by an average of 4.7{\%} semantic F1 across five datasets. Our collective findings pave the way for deeper future investigations into PLM-based KPG.

Rethinking Model Selection and Decoding for Keyphrase Generation with Pre-trained Sequence-to-Sequence Models

emnlp-main.410

2310.06374

https://huggingface.co/papers/2310.06374

Poster

https://aclanthology.org/2023.emnlp-main.411.bib

https://aclanthology.org/2023.emnlp-main.411/

@inproceedings{bast-etal-2023-fair, title = "A Fair and In-Depth Evaluation of Existing End-to-End Entity Linking Systems", author = "Bast, Hannah and Hertel, Matthias and Prange, Natalie", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.411", doi = "10.18653/v1/2023.emnlp-main.411", pages = "6659--6672", abstract = "Existing evaluations of entity linking systems often say little about how the system is going to perform for a particular application. There are two fundamental reasons for this. One is that many evaluations only use aggregate measures (like precision, recall, and F1 score), without a detailed error analysis or a closer look at the results. The other is that all of the widely used benchmarks have strong biases and artifacts, in particular: a strong focus on named entities, an unclear or missing specification of what else counts as an entity mention, poor handling of ambiguities, and an over- or underrepresentation of certain kinds of entities. We provide a more meaningful and fair in-depth evaluation of a variety of existing end-to-end entity linkers. We characterize their strengths and weaknesses and also report on reproducibility aspects. The detailed results of our evaluation can be inspected under https://elevant.cs.uni-freiburg.de/emnlp2023. Our evaluation is based on several widely used benchmarks, which exhibit the problems mentioned above to various degrees, as well as on two new benchmarks, which address the problems mentioned above. The new benchmarks can be found under https://github.com/ad-freiburg/fair-entity-linking-benchmarks.", }

Existing evaluations of entity linking systems often say little about how the system is going to perform for a particular application. There are two fundamental reasons for this. One is that many evaluations only use aggregate measures (like precision, recall, and F1 score), without a detailed error analysis or a closer look at the results. The other is that all of the widely used benchmarks have strong biases and artifacts, in particular: a strong focus on named entities, an unclear or missing specification of what else counts as an entity mention, poor handling of ambiguities, and an over- or underrepresentation of certain kinds of entities. We provide a more meaningful and fair in-depth evaluation of a variety of existing end-to-end entity linkers. We characterize their strengths and weaknesses and also report on reproducibility aspects. The detailed results of our evaluation can be inspected under https://elevant.cs.uni-freiburg.de/emnlp2023. Our evaluation is based on several widely used benchmarks, which exhibit the problems mentioned above to various degrees, as well as on two new benchmarks, which address the problems mentioned above. The new benchmarks can be found under https://github.com/ad-freiburg/fair-entity-linking-benchmarks.

A Fair and In-Depth Evaluation of Existing End-to-End Entity Linking Systems

emnlp-main.411

2305.14937

Poster

https://aclanthology.org/2023.emnlp-main.412.bib

https://aclanthology.org/2023.emnlp-main.412/

@inproceedings{wu-etal-2023-multi-task, title = "A Multi-Task Dataset for Assessing Discourse Coherence in {C}hinese Essays: Structure, Theme, and Logic Analysis", author = "Wu, Hongyi and Shen, Xinshu and Lan, Man and Mao, Shaoguang and Bai, Xiaopeng and Wu, Yuanbin", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.412", doi = "10.18653/v1/2023.emnlp-main.412", pages = "6673--6688", abstract = "This paper introduces the \textbf{C}hinese \textbf{E}ssay \textbf{D}iscourse \textbf{C}oherence \textbf{C}orpus (\textbf{CEDCC}), a multi-task dataset for assessing discourse coherence. Existing research tends to focus on isolated dimensions of discourse coherence, a gap which the CEDCC addresses by integrating coherence grading, topical continuity, and discourse relations. This approach, alongside detailed annotations, captures the subtleties of real-world texts and stimulates progress in Chinese discourse coherence analysis. Our contributions include the development of the CEDCC, the establishment of baselines for further research, and the demonstration of the impact of coherence on discourse relation recognition and automated essay scoring. The dataset and related codes is available at \url{https://github.com/cubenlp/CEDCC_corpus}.", }

This paper introduces the \textbf{C}hinese \textbf{E}ssay \textbf{D}iscourse \textbf{C}oherence \textbf{C}orpus (\textbf{CEDCC}), a multi-task dataset for assessing discourse coherence. Existing research tends to focus on isolated dimensions of discourse coherence, a gap which the CEDCC addresses by integrating coherence grading, topical continuity, and discourse relations. This approach, alongside detailed annotations, captures the subtleties of real-world texts and stimulates progress in Chinese discourse coherence analysis. Our contributions include the development of the CEDCC, the establishment of baselines for further research, and the demonstration of the impact of coherence on discourse relation recognition and automated essay scoring. The dataset and related codes is available at \url{https://github.com/cubenlp/CEDCC_corpus}.

A Multi-Task Dataset for Assessing Discourse Coherence in Chinese Essays: Structure, Theme, and Logic Analysis

emnlp-main.412

Poster

https://aclanthology.org/2023.emnlp-main.413.bib

https://aclanthology.org/2023.emnlp-main.413/

@inproceedings{chen-he-2023-skd, title = "{SKD}-{NER}: Continual Named Entity Recognition via Span-based Knowledge Distillation with Reinforcement Learning", author = "Chen, Yi and He, Liang", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.413", doi = "10.18653/v1/2023.emnlp-main.413", pages = "6689--6700", abstract = "Continual learning for named entity recognition (CL-NER) aims to enable models to continuously learn new entity types while retaining the ability to recognize previously learned ones. However, the current strategies fall short of effectively addressing the catastrophic forgetting of previously learned entity types. To tackle this issue, we propose the SKD-NER model, an efficient continual learning NER model based on the span-based approach, which innovatively incorporates reinforcement learning strategies to enhance the model{'}s ability against catastrophic forgetting. Specifically, we leverage knowledge distillation (KD) to retain memory and employ reinforcement learning strategies during the KD process to optimize the soft labeling and distillation losses generated by the teacher model to effectively prevent catastrophic forgetting during continual learning. This approach effectively prevents or mitigates catastrophic forgetting during continuous learning, allowing the model to retain previously learned knowledge while acquiring new knowledge. Our experiments on two benchmark datasets demonstrate that our model significantly improves the performance of the CL-NER task, outperforming state-of-the-art methods.", }

Continual learning for named entity recognition (CL-NER) aims to enable models to continuously learn new entity types while retaining the ability to recognize previously learned ones. However, the current strategies fall short of effectively addressing the catastrophic forgetting of previously learned entity types. To tackle this issue, we propose the SKD-NER model, an efficient continual learning NER model based on the span-based approach, which innovatively incorporates reinforcement learning strategies to enhance the model{'}s ability against catastrophic forgetting. Specifically, we leverage knowledge distillation (KD) to retain memory and employ reinforcement learning strategies during the KD process to optimize the soft labeling and distillation losses generated by the teacher model to effectively prevent catastrophic forgetting during continual learning. This approach effectively prevents or mitigates catastrophic forgetting during continuous learning, allowing the model to retain previously learned knowledge while acquiring new knowledge. Our experiments on two benchmark datasets demonstrate that our model significantly improves the performance of the CL-NER task, outperforming state-of-the-art methods.

SKD-NER: Continual Named Entity Recognition via Span-based Knowledge Distillation with Reinforcement Learning

emnlp-main.413

Poster

https://aclanthology.org/2023.emnlp-main.414.bib

https://aclanthology.org/2023.emnlp-main.414/

@inproceedings{qin-etal-2023-lifelong, title = "Lifelong Sequence Generation with Dynamic Module Expansion and Adaptation", author = "Qin, Chengwei and Chen, Chen and Joty, Shafiq", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.414", doi = "10.18653/v1/2023.emnlp-main.414", pages = "6701--6714", abstract = "Lifelong sequence generation (LSG), a problem in continual learning, aims to continually train a model on a sequence of generation tasks to learn constantly emerging new generation patterns while avoiding the forgetting of previous knowledge. Existing LSG methods mainly focus on maintaining old knowledge while paying little attention to knowledge transfer across tasks. In contrast, humans can better learn new tasks by leveraging previously acquired knowledge from similar tasks. Inspired by the learning paradigm of humans, we propose Dynamic Module Expansion and Adaptation (DMEA), which enables the model to dynamically determine the architecture for acquiring new knowledge based on task correlation and select the most similar previous tasks to facilitate adaptation to new tasks. In addition, as the learning process can easily be biased towards the current task which might cause more severe forgetting of previously learned knowledge, we propose dynamic gradient scaling to balance the learning of the current task and replayed tasks. With extensive experiments, we demonstrate that DMEA can consistently outperform existing methods in different LSG settings.", }

Lifelong sequence generation (LSG), a problem in continual learning, aims to continually train a model on a sequence of generation tasks to learn constantly emerging new generation patterns while avoiding the forgetting of previous knowledge. Existing LSG methods mainly focus on maintaining old knowledge while paying little attention to knowledge transfer across tasks. In contrast, humans can better learn new tasks by leveraging previously acquired knowledge from similar tasks. Inspired by the learning paradigm of humans, we propose Dynamic Module Expansion and Adaptation (DMEA), which enables the model to dynamically determine the architecture for acquiring new knowledge based on task correlation and select the most similar previous tasks to facilitate adaptation to new tasks. In addition, as the learning process can easily be biased towards the current task which might cause more severe forgetting of previously learned knowledge, we propose dynamic gradient scaling to balance the learning of the current task and replayed tasks. With extensive experiments, we demonstrate that DMEA can consistently outperform existing methods in different LSG settings.

Lifelong Sequence Generation with Dynamic Module Expansion and Adaptation

emnlp-main.414

2310.09886

Poster

https://aclanthology.org/2023.emnlp-main.415.bib

https://aclanthology.org/2023.emnlp-main.415/

@inproceedings{fleisig-etal-2023-majority, title = "When the Majority is Wrong: Modeling Annotator Disagreement for Subjective Tasks", author = "Fleisig, Eve and Abebe, Rediet and Klein, Dan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.415", doi = "10.18653/v1/2023.emnlp-main.415", pages = "6715--6726", abstract = "Though majority vote among annotators is typically used for ground truth labels in machine learning, annotator disagreement in tasks such as hate speech detection may reflect systematic differences in opinion across groups, not noise. Thus, a crucial problem in hate speech detection is determining if a statement is offensive to the demographic group that it targets, when that group may be a small fraction of the annotator pool. We construct a model that predicts individual annotator ratings on potentially offensive text and combines this information with the predicted target group of the text to predict the ratings of target group members. We show gains across a range of metrics, including raising performance over the baseline by 22{\%} at predicting individual annotators{'} ratings and by 33{\%} at predicting variance among annotators, which provides a metric for model uncertainty downstream. We find that annotators{'} ratings can be predicted using their demographic information as well as opinions on online content, and that non-invasive questions on annotators{'} online experiences minimize the need to collect demographic information when predicting annotators{'} opinions.", }

Though majority vote among annotators is typically used for ground truth labels in machine learning, annotator disagreement in tasks such as hate speech detection may reflect systematic differences in opinion across groups, not noise. Thus, a crucial problem in hate speech detection is determining if a statement is offensive to the demographic group that it targets, when that group may be a small fraction of the annotator pool. We construct a model that predicts individual annotator ratings on potentially offensive text and combines this information with the predicted target group of the text to predict the ratings of target group members. We show gains across a range of metrics, including raising performance over the baseline by 22{\%} at predicting individual annotators{'} ratings and by 33{\%} at predicting variance among annotators, which provides a metric for model uncertainty downstream. We find that annotators{'} ratings can be predicted using their demographic information as well as opinions on online content, and that non-invasive questions on annotators{'} online experiences minimize the need to collect demographic information when predicting annotators{'} opinions.

When the Majority is Wrong: Modeling Annotator Disagreement for Subjective Tasks

emnlp-main.415

2305.06626

Poster

https://aclanthology.org/2023.emnlp-main.416.bib

https://aclanthology.org/2023.emnlp-main.416/

@inproceedings{hemmer-etal-2023-lazy, title = "Lazy-k Decoding: Constrained Decoding for Information Extraction", author = "Hemmer, Arthur and Coustaty, Mickael and Bartolo, Nicola and Brachat, Jerome and Ogier, Jean-marc", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.416", doi = "10.18653/v1/2023.emnlp-main.416", pages = "6727--6736", abstract = "We explore the possibility of improving probabilistic models in structured prediction. Specifically, we combine the models with constrained decoding approaches in the context of token classification for information extraction. The decoding methods search for constraint-satisfying label-assignments while maximizing the total probability. To do this, we evaluate several existing approaches, as well as propose a novel decoding method called Lazy-$k$. Our findings demonstrate that constrained decoding approaches can significantly improve the models{'} performances, especially when using smaller models. The Lazy-$k$ approach allows for more flexibility between decoding time and accuracy. The code for using Lazy-$k$ decoding can be found at https://github.com/ArthurDevNL/lazyk.", }

We explore the possibility of improving probabilistic models in structured prediction. Specifically, we combine the models with constrained decoding approaches in the context of token classification for information extraction. The decoding methods search for constraint-satisfying label-assignments while maximizing the total probability. To do this, we evaluate several existing approaches, as well as propose a novel decoding method called Lazy-$k$. Our findings demonstrate that constrained decoding approaches can significantly improve the models{'} performances, especially when using smaller models. The Lazy-$k$ approach allows for more flexibility between decoding time and accuracy. The code for using Lazy-$k$ decoding can be found at https://github.com/ArthurDevNL/lazyk.

Lazy-k Decoding: Constrained Decoding for Information Extraction

emnlp-main.416

Poster

https://aclanthology.org/2023.emnlp-main.417.bib

https://aclanthology.org/2023.emnlp-main.417/

@inproceedings{chen-etal-2023-personalized, title = "Personalized Distillation: Empowering Open-Sourced {LLM}s with Adaptive Learning for Code Generation", author = "Chen, Hailin and Saha, Amrita and Hoi, Steven and Joty, Shafiq", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.417", doi = "10.18653/v1/2023.emnlp-main.417", pages = "6737--6749", abstract = "With the rise of powerful closed-sourced LLMs (ChatGPT, GPT-4), there are increasing interests in distilling the capabilies of close-sourced LLMs to smaller open-sourced LLMs. Previous distillation methods usually prompt ChatGPT to generate a set of instructions and answers, for the student model to learn. However, such standard distillation approach neglects the merits and conditions of the student model. Inspired by modern teaching principles, we design a personalised distillation process, in which the student attempts to solve a task first, then the teacher provides an adaptive refinement for the student to improve. Instead of feeding the student with teacher{'}s prior, personalised distillation enables personalised learning for the student model, as it only learns on examples it makes mistakes upon and learns to improve its own solution. On code generation, personalised distillation consistently outperforms standard distillation with only one third of the data. With only 2.5-3K personalised examples that incur a data-collection cost of 4-6{\$}, we boost CodeGen-mono-16B by 7{\%} to achieve 36.4{\%} pass@1 and StarCoder by 12.2{\%} to achieve 45.8{\%} pass@1 on HumanEval.", }

With the rise of powerful closed-sourced LLMs (ChatGPT, GPT-4), there are increasing interests in distilling the capabilies of close-sourced LLMs to smaller open-sourced LLMs. Previous distillation methods usually prompt ChatGPT to generate a set of instructions and answers, for the student model to learn. However, such standard distillation approach neglects the merits and conditions of the student model. Inspired by modern teaching principles, we design a personalised distillation process, in which the student attempts to solve a task first, then the teacher provides an adaptive refinement for the student to improve. Instead of feeding the student with teacher{'}s prior, personalised distillation enables personalised learning for the student model, as it only learns on examples it makes mistakes upon and learns to improve its own solution. On code generation, personalised distillation consistently outperforms standard distillation with only one third of the data. With only 2.5-3K personalised examples that incur a data-collection cost of 4-6{\$}, we boost CodeGen-mono-16B by 7{\%} to achieve 36.4{\%} pass@1 and StarCoder by 12.2{\%} to achieve 45.8{\%} pass@1 on HumanEval.

Personalized Distillation: Empowering Open-Sourced LLMs with Adaptive Learning for Code Generation

emnlp-main.417

Poster

https://aclanthology.org/2023.emnlp-main.418.bib

https://aclanthology.org/2023.emnlp-main.418/

@inproceedings{jain-etal-2023-language-models, title = "Do Language Models Have a Common Sense regarding Time? Revisiting Temporal Commonsense Reasoning in the Era of Large Language Models", author = "Jain, Raghav and Sojitra, Daivik and Acharya, Arkadeep and Saha, Sriparna and Jatowt, Adam and Dandapat, Sandipan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.418", doi = "10.18653/v1/2023.emnlp-main.418", pages = "6750--6774", abstract = "Temporal reasoning represents a vital component of human communication and understanding, yet remains an underexplored area within the context of Large Language Models (LLMs). Despite LLMs demonstrating significant proficiency in a range of tasks, a comprehensive, large-scale analysis of their temporal reasoning capabilities is missing. Our paper addresses this gap, presenting the first extensive benchmarking of LLMs on temporal reasoning tasks. We critically evaluate 8 different LLMs across 6 datasets using 3 distinct prompting strategies. Additionally, we broaden the scope of our evaluation by including in our analysis 2 Code Generation LMs. Beyond broad benchmarking of models and prompts, we also conduct a fine-grained investigation of performance across different categories of temporal tasks. We further analyze the LLMs on varying temporal aspects, offering insights into their proficiency in understanding and predicting the continuity, sequence, and progression of events over time. Our findings reveal a nuanced depiction of the capabilities and limitations of the models within temporal reasoning, offering a comprehensive reference for future research in this pivotal domain.", }

Temporal reasoning represents a vital component of human communication and understanding, yet remains an underexplored area within the context of Large Language Models (LLMs). Despite LLMs demonstrating significant proficiency in a range of tasks, a comprehensive, large-scale analysis of their temporal reasoning capabilities is missing. Our paper addresses this gap, presenting the first extensive benchmarking of LLMs on temporal reasoning tasks. We critically evaluate 8 different LLMs across 6 datasets using 3 distinct prompting strategies. Additionally, we broaden the scope of our evaluation by including in our analysis 2 Code Generation LMs. Beyond broad benchmarking of models and prompts, we also conduct a fine-grained investigation of performance across different categories of temporal tasks. We further analyze the LLMs on varying temporal aspects, offering insights into their proficiency in understanding and predicting the continuity, sequence, and progression of events over time. Our findings reveal a nuanced depiction of the capabilities and limitations of the models within temporal reasoning, offering a comprehensive reference for future research in this pivotal domain.

Do Language Models Have a Common Sense regarding Time? Revisiting Temporal Commonsense Reasoning in the Era of Large Language Models

emnlp-main.418

Poster

https://aclanthology.org/2023.emnlp-main.419.bib

https://aclanthology.org/2023.emnlp-main.419/

@inproceedings{havaldar-etal-2023-comparing, title = "Comparing Styles across Languages", author = "Havaldar, Shreya and Pressimone, Matthew and Wong, Eric and Ungar, Lyle", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.419", doi = "10.18653/v1/2023.emnlp-main.419", pages = "6775--6791", abstract = "Understanding how styles differ across languages is advantageous for training both humans and computers to generate culturally appropriate text. We introduce an explanation framework to extract stylistic differences from multilingual LMs and compare styles across languages. Our framework (1) generates comprehensive style lexica in any language and (2) consolidates feature importances from LMs into comparable lexical categories. We apply this framework to compare politeness, creating the first holistic multilingual politeness dataset and exploring how politeness varies across four languages. Our approach enables an effective evaluation of how distinct linguistic categories contribute to stylistic variations and provides interpretable insights into how people communicate differently around the world.", }

Understanding how styles differ across languages is advantageous for training both humans and computers to generate culturally appropriate text. We introduce an explanation framework to extract stylistic differences from multilingual LMs and compare styles across languages. Our framework (1) generates comprehensive style lexica in any language and (2) consolidates feature importances from LMs into comparable lexical categories. We apply this framework to compare politeness, creating the first holistic multilingual politeness dataset and exploring how politeness varies across four languages. Our approach enables an effective evaluation of how distinct linguistic categories contribute to stylistic variations and provides interpretable insights into how people communicate differently around the world.

Comparing Styles across Languages

emnlp-main.419

2310.07135

Poster

https://aclanthology.org/2023.emnlp-main.420.bib

https://aclanthology.org/2023.emnlp-main.420/

@inproceedings{liu-etal-2023-event, title = "Event Causality Extraction via Implicit Cause-Effect Interactions", author = "Liu, Jintao and Zhang, Zequn and Wei, Kaiwen and Guo, Zhi and Sun, Xian and Jin, Li and Li, Xiaoyu", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.420", doi = "10.18653/v1/2023.emnlp-main.420", pages = "6792--6804", abstract = "Event Causality Extraction (ECE) aims to extract the cause-effect event pairs from the given text, which requires the model to possess a strong reasoning ability to capture event causalities. However, existing works have not adequately exploited the interactions between the cause and effect event that could provide crucial clues for causality reasoning. To this end, we propose an Implicit Cause-Effect interaction (ICE) framework, which formulates ECE as a template-based conditional generation problem. The proposed method captures the implicit intra- and inter-event interactions by incorporating the privileged information (ground truth event types and arguments) for reasoning, and a knowledge distillation mechanism is introduced to alleviate the unavailability of privileged information in the test stage. Furthermore, to facilitate knowledge transfer from teacher to student, we design an event-level alignment strategy named Cause-Effect Optimal Transport (CEOT) to strengthen the semantic interactions of cause-effect event types and arguments. Experimental results indicate that ICE achieves state-of-the-art performance on the ECE-CCKS dataset.", }

Event Causality Extraction (ECE) aims to extract the cause-effect event pairs from the given text, which requires the model to possess a strong reasoning ability to capture event causalities. However, existing works have not adequately exploited the interactions between the cause and effect event that could provide crucial clues for causality reasoning. To this end, we propose an Implicit Cause-Effect interaction (ICE) framework, which formulates ECE as a template-based conditional generation problem. The proposed method captures the implicit intra- and inter-event interactions by incorporating the privileged information (ground truth event types and arguments) for reasoning, and a knowledge distillation mechanism is introduced to alleviate the unavailability of privileged information in the test stage. Furthermore, to facilitate knowledge transfer from teacher to student, we design an event-level alignment strategy named Cause-Effect Optimal Transport (CEOT) to strengthen the semantic interactions of cause-effect event types and arguments. Experimental results indicate that ICE achieves state-of-the-art performance on the ECE-CCKS dataset.

Event Causality Extraction via Implicit Cause-Effect Interactions

emnlp-main.420

Poster

https://aclanthology.org/2023.emnlp-main.421.bib

https://aclanthology.org/2023.emnlp-main.421/

@inproceedings{deas-etal-2023-evaluation, title = "Evaluation of {A}frican {A}merican Language Bias in Natural Language Generation", author = "Deas, Nicholas and Grieser, Jessica and Kleiner, Shana and Patton, Desmond and Turcan, Elsbeth and McKeown, Kathleen", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.421", doi = "10.18653/v1/2023.emnlp-main.421", pages = "6805--6824", abstract = "While biases disadvantaging African American Language (AAL) have been uncovered in models for tasks such as speech recognition and toxicity detection, there has been little investigation of these biases for language generation models like ChatGPT. We evaluate how well LLMs understand AAL in comparison to White Mainstream English (WME), the encouraged {``}standard{''} form of English taught in American classrooms. We measure large language model performance on two tasks: a counterpart generation task, where a model generates AAL given WME and vice versa, and a masked span prediction (MSP) task, where models predict a phrase hidden from their input. Using a novel dataset of AAL texts from a variety of regions and contexts, we present evidence of dialectal bias for six pre-trained LLMs through performance gaps on these tasks.", }

While biases disadvantaging African American Language (AAL) have been uncovered in models for tasks such as speech recognition and toxicity detection, there has been little investigation of these biases for language generation models like ChatGPT. We evaluate how well LLMs understand AAL in comparison to White Mainstream English (WME), the encouraged {``}standard{''} form of English taught in American classrooms. We measure large language model performance on two tasks: a counterpart generation task, where a model generates AAL given WME and vice versa, and a masked span prediction (MSP) task, where models predict a phrase hidden from their input. Using a novel dataset of AAL texts from a variety of regions and contexts, we present evidence of dialectal bias for six pre-trained LLMs through performance gaps on these tasks.

Evaluation of African American Language Bias in Natural Language Generation

emnlp-main.421

2305.14291

Poster

https://aclanthology.org/2023.emnlp-main.422.bib

https://aclanthology.org/2023.emnlp-main.422/

@inproceedings{hu-etal-2023-systematic, title = "A Systematic Study of Performance Disparities in Multilingual Task-Oriented Dialogue Systems", author = "Hu, Songbo and Zhou, Han and Yuan, Moy and Gritta, Milan and Zhang, Guchun and Iacobacci, Ignacio and Korhonen, Anna and Vuli{\'c}, Ivan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.422", doi = "10.18653/v1/2023.emnlp-main.422", pages = "6825--6851", abstract = "Achieving robust language technologies that can perform well across the world{'}s many languages is a central goal of multilingual NLP. In this work, we take stock of and empirically analyse task performance disparities that exist between multilingual task-oriented dialogue (ToD) systems. We first define new quantitative measures of absolute and relative equivalence in system performance, capturing disparities across languages and within individual languages. Through a series of controlled experiments, we demonstrate that performance disparities depend on a number of factors: the nature of the ToD task at hand, the underlying pretrained language model, the target language, and the amount of ToD annotated data. We empirically prove the existence of the adaptation and intrinsic biases in current ToD systems: e.g., ToD systems trained for Arabic or Turkish using annotated ToD data fully parallel to English ToD data still exhibit diminished ToD task performance. Beyond providing a series of insights into the performance disparities of ToD systems in different languages, our analyses offer practical tips on how to approach ToD data collection and system development for new languages.", }

Achieving robust language technologies that can perform well across the world{'}s many languages is a central goal of multilingual NLP. In this work, we take stock of and empirically analyse task performance disparities that exist between multilingual task-oriented dialogue (ToD) systems. We first define new quantitative measures of absolute and relative equivalence in system performance, capturing disparities across languages and within individual languages. Through a series of controlled experiments, we demonstrate that performance disparities depend on a number of factors: the nature of the ToD task at hand, the underlying pretrained language model, the target language, and the amount of ToD annotated data. We empirically prove the existence of the adaptation and intrinsic biases in current ToD systems: e.g., ToD systems trained for Arabic or Turkish using annotated ToD data fully parallel to English ToD data still exhibit diminished ToD task performance. Beyond providing a series of insights into the performance disparities of ToD systems in different languages, our analyses offer practical tips on how to approach ToD data collection and system development for new languages.

A Systematic Study of Performance Disparities in Multilingual Task-Oriented Dialogue Systems

emnlp-main.422

2310.12892

https://huggingface.co/papers/2310.12892

Poster

https://aclanthology.org/2023.emnlp-main.423.bib

https://aclanthology.org/2023.emnlp-main.423/

@inproceedings{akavarapu-bhattacharya-2023-cognate, title = "Cognate Transformer for Automated Phonological Reconstruction and Cognate Reflex Prediction", author = "Akavarapu, V.S.D.S.Mahesh and Bhattacharya, Arnab", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.423", doi = "10.18653/v1/2023.emnlp-main.423", pages = "6852--6862", abstract = "Phonological reconstruction is one of the central problems in historical linguistics where a proto-word of an ancestral language is determined from the observed cognate words of daughter languages. Computational approaches to historical linguistics attempt to automate the task by learning models on available linguistic data. Several ideas and techniques drawn from computational biology have been successfully applied in this area of computational historical linguistics. Following these lines, we adapt MSA Transformer, a protein language model, to the problem of automated phonological reconstruction. MSA Transformer trains on multiple sequence alignments as input and is, thus, apt for application on aligned cognate words. We, hence, name our model as Cognate Transformer. We also apply the model on another associated task, namely, cognate reflex prediction where a reflex word in a daughter language is predicted based on cognate words from other daughter languages. We show that our model outperforms the existing models on both the tasks, especially when it is pre-trained on masked word prediction task.", }

Phonological reconstruction is one of the central problems in historical linguistics where a proto-word of an ancestral language is determined from the observed cognate words of daughter languages. Computational approaches to historical linguistics attempt to automate the task by learning models on available linguistic data. Several ideas and techniques drawn from computational biology have been successfully applied in this area of computational historical linguistics. Following these lines, we adapt MSA Transformer, a protein language model, to the problem of automated phonological reconstruction. MSA Transformer trains on multiple sequence alignments as input and is, thus, apt for application on aligned cognate words. We, hence, name our model as Cognate Transformer. We also apply the model on another associated task, namely, cognate reflex prediction where a reflex word in a daughter language is predicted based on cognate words from other daughter languages. We show that our model outperforms the existing models on both the tasks, especially when it is pre-trained on masked word prediction task.

Cognate Transformer for Automated Phonological Reconstruction and Cognate Reflex Prediction

emnlp-main.423

2310.07487

Oral

https://aclanthology.org/2023.emnlp-main.424.bib

https://aclanthology.org/2023.emnlp-main.424/

@inproceedings{lu-etal-2023-inference, title = "Inference-Time Policy Adapters ({IPA}): Tailoring Extreme-Scale {LM}s without Fine-tuning", author = "Lu, Ximing and Brahman, Faeze and West, Peter and Jung, Jaehun and Chandu, Khyathi and Ravichander, Abhilasha and Ammanabrolu, Prithviraj and Jiang, Liwei and Ramnath, Sahana and Dziri, Nouha and Fisher, Jillian and Lin, Bill and Hallinan, Skyler and Qin, Lianhui and Ren, Xiang and Welleck, Sean and Choi, Yejin", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.424", doi = "10.18653/v1/2023.emnlp-main.424", pages = "6863--6883", abstract = "While extreme-scale language models have demonstrated exceptional performance on a variety of language tasks, the degree of control over these language models through pure prompting can often be limited. Directly fine-tuning such language models can be effective for tailoring them, but it can be either extremely costly (e.g., GPT-3) or not even feasible for the broader community (e.g., GPT-4). We propose Inference-time Policy Adapters (IPA), which efficiently tailors a language model such as GPT-3 without fine-tuning it. IPA guides a large base model during decoding time through a lightweight policy adapter trained to optimize an arbitrary user objective with reinforcement learning. On five challenging text generation tasks, such as toxicity reduction and lexically constrained generation, IPA consistently brings significant improvements over off-the-shelf language models. It outperforms competitive baseline methods, sometimes even including expensive fine-tuning. In particular, tailoring GPT-2 with IPA can outperform GPT-3, while tailoring GPT-3 with IPA brings a major performance boost over GPT-3 (and sometimes even over GPT-4). Our promising results highlight the potential of IPA as a lightweight alternative to tailoring extreme-scale language models.", }

While extreme-scale language models have demonstrated exceptional performance on a variety of language tasks, the degree of control over these language models through pure prompting can often be limited. Directly fine-tuning such language models can be effective for tailoring them, but it can be either extremely costly (e.g., GPT-3) or not even feasible for the broader community (e.g., GPT-4). We propose Inference-time Policy Adapters (IPA), which efficiently tailors a language model such as GPT-3 without fine-tuning it. IPA guides a large base model during decoding time through a lightweight policy adapter trained to optimize an arbitrary user objective with reinforcement learning. On five challenging text generation tasks, such as toxicity reduction and lexically constrained generation, IPA consistently brings significant improvements over off-the-shelf language models. It outperforms competitive baseline methods, sometimes even including expensive fine-tuning. In particular, tailoring GPT-2 with IPA can outperform GPT-3, while tailoring GPT-3 with IPA brings a major performance boost over GPT-3 (and sometimes even over GPT-4). Our promising results highlight the potential of IPA as a lightweight alternative to tailoring extreme-scale language models.

Inference-Time Policy Adapters (IPA): Tailoring Extreme-Scale LMs without Fine-tuning

emnlp-main.424

2305.15065

https://huggingface.co/papers/2305.15065

Poster

https://aclanthology.org/2023.emnlp-main.425.bib

https://aclanthology.org/2023.emnlp-main.425/

@inproceedings{lee-etal-2023-weakly, title = "Weakly Supervised Semantic Parsing with Execution-based Spurious Program Filtering", author = "Lee, Kang-il and Kim, Segwang and Jung, Kyomin", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.425", doi = "10.18653/v1/2023.emnlp-main.425", pages = "6884--6894", abstract = "The problem of spurious programs is a longstanding challenge when training a semantic parser from weak supervision. To eliminate such programs that have wrong semantics but correct denotation, existing methods focus on exploiting similarities between examples based on domain-specific knowledge. In this paper, we propose a domain-agnostic filtering mechanism based on program execution results. Specifically, for each program obtained through the search process, we first construct a representation that captures the program{'}s semantics as execution results under various inputs. Then, we run a majority vote on these representations to identify and filter out programs with significantly different semantics from the other programs. In particular, our method is orthogonal to the program search process so that it can easily augment any of the existing weakly supervised semantic parsing frameworks. Empirical evaluations on the Natural Language Visual Reasoning and WikiTableQuestions demonstrate that applying our method to the existing semantic parsers induces significantly improved performances.", }

The problem of spurious programs is a longstanding challenge when training a semantic parser from weak supervision. To eliminate such programs that have wrong semantics but correct denotation, existing methods focus on exploiting similarities between examples based on domain-specific knowledge. In this paper, we propose a domain-agnostic filtering mechanism based on program execution results. Specifically, for each program obtained through the search process, we first construct a representation that captures the program{'}s semantics as execution results under various inputs. Then, we run a majority vote on these representations to identify and filter out programs with significantly different semantics from the other programs. In particular, our method is orthogonal to the program search process so that it can easily augment any of the existing weakly supervised semantic parsing frameworks. Empirical evaluations on the Natural Language Visual Reasoning and WikiTableQuestions demonstrate that applying our method to the existing semantic parsers induces significantly improved performances.

Weakly Supervised Semantic Parsing with Execution-based Spurious Program Filtering

emnlp-main.425

2311.01161

Poster

https://aclanthology.org/2023.emnlp-main.426.bib

https://aclanthology.org/2023.emnlp-main.426/

@inproceedings{k-etal-2023-taxonomy, title = "Taxonomy Expansion for Named Entity Recognition", author = "K, Karthikeyan and Vyas, Yogarshi and Ma, Jie and Paolini, Giovanni and John, Neha and Wang, Shuai and Benajiba, Yassine and Castelli, Vittorio and Roth, Dan and Ballesteros, Miguel", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.426", doi = "10.18653/v1/2023.emnlp-main.426", pages = "6895--6906", abstract = "Training a Named Entity Recognition (NER) model often involves fixing a taxonomy of entity types. However, requirements evolve and we might need the NER model to recognize additional entity types. A simple approach is to re-annotate entire dataset with both existing and additional entity types and then train the model on the re-annotated dataset. However, this is an extremely laborious task. To remedy this, we propose a novel approach called Partial Label Model (PLM) that uses only partially annotated datasets. We experiment with 6 diverse datasets and show that PLM consistently performs better than most other approaches (0.5 - 2.5 F1), including in novel settings for taxonomy expansion not considered in prior work. The gap between PLM and all other approaches is especially large in settings where there is limited data available for the additional entity types (as much as 11 F1), thus suggesting a more cost effective approaches to taxonomy expansion.", }

Training a Named Entity Recognition (NER) model often involves fixing a taxonomy of entity types. However, requirements evolve and we might need the NER model to recognize additional entity types. A simple approach is to re-annotate entire dataset with both existing and additional entity types and then train the model on the re-annotated dataset. However, this is an extremely laborious task. To remedy this, we propose a novel approach called Partial Label Model (PLM) that uses only partially annotated datasets. We experiment with 6 diverse datasets and show that PLM consistently performs better than most other approaches (0.5 - 2.5 F1), including in novel settings for taxonomy expansion not considered in prior work. The gap between PLM and all other approaches is especially large in settings where there is limited data available for the additional entity types (as much as 11 F1), thus suggesting a more cost effective approaches to taxonomy expansion.

Taxonomy Expansion for Named Entity Recognition

emnlp-main.426

2305.13191

Poster

https://aclanthology.org/2023.emnlp-main.427.bib

https://aclanthology.org/2023.emnlp-main.427/

@inproceedings{eberle-etal-2023-rather, title = "Rather a Nurse than a Physician - Contrastive Explanations under Investigation", author = "Eberle, Oliver and Chalkidis, Ilias and Cabello, Laura and Brandl, Stephanie", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.427", doi = "10.18653/v1/2023.emnlp-main.427", pages = "6907--6920", abstract = "Contrastive explanations, where one decision is explained *in contrast to another*, are supposed to be closer to how humans explain a decision than non-contrastive explanations, where the decision is not necessarily referenced to an alternative. This claim has never been empirically validated. We analyze four English text-classification datasets (SST2, DynaSent, BIOS and DBpedia-Animals). We fine-tune and extract explanations from three different models (RoBERTa, GTP-2, and T5), each in three different sizes and apply three post-hoc explainability methods (LRP, GradientxInput, GradNorm). We furthermore collect and release human rationale annotations for a subset of 100 samples from the BIOS dataset for contrastive and non-contrastive settings. A cross-comparison between model-based rationales and human annotations, both in contrastive and non-contrastive settings, yields a high agreement between the two settings for models as well as for humans. Moreover, model-based explanations computed in both settings align equally well with human rationales. Thus, we empirically find that humans do not necessarily explain in a contrastive manner.", }

Contrastive explanations, where one decision is explained *in contrast to another*, are supposed to be closer to how humans explain a decision than non-contrastive explanations, where the decision is not necessarily referenced to an alternative. This claim has never been empirically validated. We analyze four English text-classification datasets (SST2, DynaSent, BIOS and DBpedia-Animals). We fine-tune and extract explanations from three different models (RoBERTa, GTP-2, and T5), each in three different sizes and apply three post-hoc explainability methods (LRP, GradientxInput, GradNorm). We furthermore collect and release human rationale annotations for a subset of 100 samples from the BIOS dataset for contrastive and non-contrastive settings. A cross-comparison between model-based rationales and human annotations, both in contrastive and non-contrastive settings, yields a high agreement between the two settings for models as well as for humans. Moreover, model-based explanations computed in both settings align equally well with human rationales. Thus, we empirically find that humans do not necessarily explain in a contrastive manner.

Rather a Nurse than a Physician - Contrastive Explanations under Investigation

emnlp-main.427

2310.11906

https://huggingface.co/papers/2310.11906

Poster

https://aclanthology.org/2023.emnlp-main.428.bib

https://aclanthology.org/2023.emnlp-main.428/

@inproceedings{dwivedi-etal-2023-eticor, title = "{E}ti{C}or: Corpus for Analyzing {LLM}s for Etiquettes", author = "Dwivedi, Ashutosh and Lavania, Pradhyumna and Modi, Ashutosh", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.428", doi = "10.18653/v1/2023.emnlp-main.428", pages = "6921--6931", abstract = "Etiquettes are an essential ingredient of day-to-day interactions among people. Moreover, etiquettes are region-specific, and etiquettes in one region might contradict those in other regions. In this paper, we propose EtiCor, an Etiquettes Corpus, having texts about social norms from five different regions across the globe. The corpus provides a test bed for evaluating LLMs for knowledge and understanding of region-specific etiquettes. Additionally, we propose the task of Etiquette Sensitivity. We experiment with state-of-the-art LLMs (Delphi, Falcon40B, and GPT-3.5). Initial results indicate that LLMs, mostly fail to understand etiquettes from regions from non-Western world.", }

Etiquettes are an essential ingredient of day-to-day interactions among people. Moreover, etiquettes are region-specific, and etiquettes in one region might contradict those in other regions. In this paper, we propose EtiCor, an Etiquettes Corpus, having texts about social norms from five different regions across the globe. The corpus provides a test bed for evaluating LLMs for knowledge and understanding of region-specific etiquettes. Additionally, we propose the task of Etiquette Sensitivity. We experiment with state-of-the-art LLMs (Delphi, Falcon40B, and GPT-3.5). Initial results indicate that LLMs, mostly fail to understand etiquettes from regions from non-Western world.

EtiCor: Corpus for Analyzing LLMs for Etiquettes

emnlp-main.428

2310.18974

Poster

https://aclanthology.org/2023.emnlp-main.429.bib

https://aclanthology.org/2023.emnlp-main.429/

@inproceedings{qi-etal-2023-investigation, title = "An Investigation of {LLM}s{'} Inefficacy in Understanding Converse Relations", author = "Qi, Chengwen and Li, Bowen and Hui, Binyuan and Wang, Bailin and Li, Jinyang and Wu, Jinwang and Laili, Yuanjun", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.429", pages = "6932--6953", abstract = "Large Language Models (LLMs) have achieved remarkable success in many formal language oriented tasks, such as structural data-to-text and semantic parsing. However current benchmarks mostly follow the data distribution of the pre-training data of LLMs. Therefore, a natural question rises that do LLMs really understand the structured semantics of formal languages. In this paper, we investigate this problem on a special case, converse binary relation. We introduce a new benchmark ConvRe focusing on converse relations, which contains 17 relations and 1240 triples extracted from popular knowledge graph completion datasets. Our ConvRE features two tasks, Re2Text and Text2Re, which are formulated as multi-choice question answering to evaluate LLMs{'} ability to determine the matching between relations and associated text. For the evaluation protocol, apart from different prompting methods, we further introduce variants to the test text and few-shot example text. We conduct experiments on three popular LLM families and have observed various scaling trends. The results suggest that LLMs often resort to shortcut learning and still face challenges on our proposed benchmark.", }

Large Language Models (LLMs) have achieved remarkable success in many formal language oriented tasks, such as structural data-to-text and semantic parsing. However current benchmarks mostly follow the data distribution of the pre-training data of LLMs. Therefore, a natural question rises that do LLMs really understand the structured semantics of formal languages. In this paper, we investigate this problem on a special case, converse binary relation. We introduce a new benchmark ConvRe focusing on converse relations, which contains 17 relations and 1240 triples extracted from popular knowledge graph completion datasets. Our ConvRE features two tasks, Re2Text and Text2Re, which are formulated as multi-choice question answering to evaluate LLMs{'} ability to determine the matching between relations and associated text. For the evaluation protocol, apart from different prompting methods, we further introduce variants to the test text and few-shot example text. We conduct experiments on three popular LLM families and have observed various scaling trends. The results suggest that LLMs often resort to shortcut learning and still face challenges on our proposed benchmark.

An Investigation of LLMs' Inefficacy in Understanding Converse Relations

emnlp-main.429

2310.05163

https://huggingface.co/papers/2310.05163

Poster

https://aclanthology.org/2023.emnlp-main.430.bib

https://aclanthology.org/2023.emnlp-main.430/

@inproceedings{wang-etal-2023-towards-low, title = "Towards Low-Resource Automatic Program Repair with Meta-Learning and Pretrained Language Models", author = "Wang, Weishi and Wang, Yue and Hoi, Steven and Joty, Shafiq", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.430", doi = "10.18653/v1/2023.emnlp-main.430", pages = "6954--6968", abstract = "Automatic program repair (APR) has gained increasing attention as an essential technique in software development to reduce manual debugging efforts and boost developers{'} productivity. Recent advances in deep learning (DL) based models have demonstrated promising results by learning from large-scale bug-fix examples in a data-driven manner. However, in practical scenarios, software bugs have an imbalanced distribution, and the fixing knowledge learned by APR models often only capture the patterns of frequent error types, making it inapplicable to handle the rare error types. To address this limitation, we investigate a novel task of low-resource APR, and propose Meta-APR, a new meta-learning framework integrated with code pretrained language models to generate fixes for low-resource bugs with limited training samples. Our Meta-APR learns better error-specific knowledge from high-resource bugs through efficient first-order meta-learning optimization, which allows for a faster adaptation to the target low-resource bugs. Besides, while we adopt CodeT5, a pretrained code-aware encoder-decoder Transformer, as the backbone model for Meta-APR, it is a model-agnostic framework that can be integrated with any neural models. Extensive experimental results on three benchmarks in various programming languages verify the superiority of our method over existing DL-based APR approaches.", }

Automatic program repair (APR) has gained increasing attention as an essential technique in software development to reduce manual debugging efforts and boost developers{'} productivity. Recent advances in deep learning (DL) based models have demonstrated promising results by learning from large-scale bug-fix examples in a data-driven manner. However, in practical scenarios, software bugs have an imbalanced distribution, and the fixing knowledge learned by APR models often only capture the patterns of frequent error types, making it inapplicable to handle the rare error types. To address this limitation, we investigate a novel task of low-resource APR, and propose Meta-APR, a new meta-learning framework integrated with code pretrained language models to generate fixes for low-resource bugs with limited training samples. Our Meta-APR learns better error-specific knowledge from high-resource bugs through efficient first-order meta-learning optimization, which allows for a faster adaptation to the target low-resource bugs. Besides, while we adopt CodeT5, a pretrained code-aware encoder-decoder Transformer, as the backbone model for Meta-APR, it is a model-agnostic framework that can be integrated with any neural models. Extensive experimental results on three benchmarks in various programming languages verify the superiority of our method over existing DL-based APR approaches.

Towards Low-Resource Automatic Program Repair with Meta-Learning and Pretrained Language Models

emnlp-main.430

Poster

https://aclanthology.org/2023.emnlp-main.431.bib

https://aclanthology.org/2023.emnlp-main.431/

@inproceedings{rathore-etal-2023-zgul, title = "{ZGUL}: Zero-shot Generalization to Unseen Languages using Multi-source Ensembling of Language Adapters", author = "Rathore, Vipul and Dhingra, Rajdeep and Singla, Parag and {Mausam}", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.431", doi = "10.18653/v1/2023.emnlp-main.431", pages = "6969--6987", abstract = "We tackle the problem of zero-shot cross-lingual transfer in NLP tasks via the use of language adapters (LAs). Most of the earlier works have explored training with adapter of a single source (often English), and testing either using the target LA or LA of another related language. Training target LA requires unlabeled data, which may not be readily available for low resource *unseen* languages: those that are neither seen by the underlying multilingual language model (e.g., mBERT), nor do we have any (labeled or unlabeled) data for them. We posit that for more effective cross-lingual transfer, instead of just one source LA, we need to leverage LAs of multiple (linguistically or geographically related) source languages, both at train and test-time - which we investigate via our novel neural architecture, ZGUL. Extensive experimentation across four language groups, covering 15 unseen target languages, demonstrates improvements of up to 3.2 average F1 points over standard fine-tuning and other strong baselines on POS tagging and NER tasks. We also extend ZGUL to settings where either (1) some unlabeled data or (2) few-shot training examples are available for the target language. We find that ZGUL continues to outperform baselines in these settings too.", }

We tackle the problem of zero-shot cross-lingual transfer in NLP tasks via the use of language adapters (LAs). Most of the earlier works have explored training with adapter of a single source (often English), and testing either using the target LA or LA of another related language. Training target LA requires unlabeled data, which may not be readily available for low resource *unseen* languages: those that are neither seen by the underlying multilingual language model (e.g., mBERT), nor do we have any (labeled or unlabeled) data for them. We posit that for more effective cross-lingual transfer, instead of just one source LA, we need to leverage LAs of multiple (linguistically or geographically related) source languages, both at train and test-time - which we investigate via our novel neural architecture, ZGUL. Extensive experimentation across four language groups, covering 15 unseen target languages, demonstrates improvements of up to 3.2 average F1 points over standard fine-tuning and other strong baselines on POS tagging and NER tasks. We also extend ZGUL to settings where either (1) some unlabeled data or (2) few-shot training examples are available for the target language. We find that ZGUL continues to outperform baselines in these settings too.

ZGUL: Zero-shot Generalization to Unseen Languages using Multi-source Ensembling of Language Adapters

emnlp-main.431

2310.16393

Poster

https://aclanthology.org/2023.emnlp-main.432.bib

https://aclanthology.org/2023.emnlp-main.432/

@inproceedings{han-etal-2023-log, title = "Log-{FGAER}: Logic-Guided Fine-Grained Address Entity Recognition from Multi-Turn Spoken Dialogue", author = "Han, Xue and Wang, Yitong and Hu, Qian and Hu, Pengwei and Deng, Chao and Feng, Junlan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.432", doi = "10.18653/v1/2023.emnlp-main.432", pages = "6988--6997", abstract = "Fine-grained address entity recognition (FGAER) from multi-turn spoken dialogues is particularly challenging. The major reason lies in that a full address is often formed through a conversation process. Different parts of an address are distributed through multiple turns of a dialogue with spoken noises. It is nontrivial to extract by turn and combine them. This challenge has not been well emphasized by main-stream entity extraction algorithms. To address this issue, we propose in this paper a logic-guided fine-grained address recognition method (Log-FGAER), where we formulate the address hierarchy relationship as the logic rule and softly apply it in a probabilistic manner to improve the accuracy of FGAER. In addition, we provide an ontology-based data augmentation methodology that employs ChatGPT to augment a spoken dialogue dataset with labeled address entities. Experiments are conducted using datasets generated by the proposed data augmentation technique and derived from real-world scenarios. The results of the experiment demonstrate the efficacy of our proposal.", }

Fine-grained address entity recognition (FGAER) from multi-turn spoken dialogues is particularly challenging. The major reason lies in that a full address is often formed through a conversation process. Different parts of an address are distributed through multiple turns of a dialogue with spoken noises. It is nontrivial to extract by turn and combine them. This challenge has not been well emphasized by main-stream entity extraction algorithms. To address this issue, we propose in this paper a logic-guided fine-grained address recognition method (Log-FGAER), where we formulate the address hierarchy relationship as the logic rule and softly apply it in a probabilistic manner to improve the accuracy of FGAER. In addition, we provide an ontology-based data augmentation methodology that employs ChatGPT to augment a spoken dialogue dataset with labeled address entities. Experiments are conducted using datasets generated by the proposed data augmentation technique and derived from real-world scenarios. The results of the experiment demonstrate the efficacy of our proposal.

Log-FGAER: Logic-Guided Fine-Grained Address Entity Recognition from Multi-Turn Spoken Dialogue

emnlp-main.432

Poster

https://aclanthology.org/2023.emnlp-main.433.bib

https://aclanthology.org/2023.emnlp-main.433/

@inproceedings{das-etal-2023-unified, title = "Unified Low-Resource Sequence Labeling by Sample-Aware Dynamic Sparse Finetuning", author = "Das, Sarkar Snigdha Sarathi and Zhang, Haoran and Shi, Peng and Yin, Wenpeng and Zhang, Rui", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.433", doi = "10.18653/v1/2023.emnlp-main.433", pages = "6998--7010", abstract = "Unified Sequence Labeling that articulates different sequence labeling problems such as Named Entity Recognition, Relation Extraction, Semantic Role Labeling, etc. in a generalized sequence-to-sequence format opens up the opportunity to make the maximum utilization of large language model knowledge toward structured prediction. Unfortunately, this requires formatting them into specialized augmented format unknown to the base pretrained language model (PLMs) necessitating finetuning to the target format. This significantly bounds its usefulness in data-limited settings where finetuning large models cannot properly generalize to the target format. To address this challenge and leverage PLM knowledge effectively, we propose FISH-DIP, a sample-aware dynamic sparse finetuning strategy that selectively focuses on a fraction of parameters, informed by feedback from highly regressing examples, during the fine-tuning process. By leveraging the dynamism of sparsity, our approach mitigates the impact of well-learned samples and prioritizes underperforming instances for improvement in generalization. Across five tasks of sequence labeling, we demonstrate that FISH-DIP can smoothly optimize the model in low resource settings offering upto 40{\%} performance improvements over full fine-tuning depending on target evaluation settings. Also, compared to in-context learning and other parameter-efficient fine-tuning approaches, FISH-DIP performs comparably or better, notably in extreme low-resource settings. The source code of FISH-DIP will be available at [this URL](https://github.com/psunlpgroup/FISH-DIP)", }

Unified Sequence Labeling that articulates different sequence labeling problems such as Named Entity Recognition, Relation Extraction, Semantic Role Labeling, etc. in a generalized sequence-to-sequence format opens up the opportunity to make the maximum utilization of large language model knowledge toward structured prediction. Unfortunately, this requires formatting them into specialized augmented format unknown to the base pretrained language model (PLMs) necessitating finetuning to the target format. This significantly bounds its usefulness in data-limited settings where finetuning large models cannot properly generalize to the target format. To address this challenge and leverage PLM knowledge effectively, we propose FISH-DIP, a sample-aware dynamic sparse finetuning strategy that selectively focuses on a fraction of parameters, informed by feedback from highly regressing examples, during the fine-tuning process. By leveraging the dynamism of sparsity, our approach mitigates the impact of well-learned samples and prioritizes underperforming instances for improvement in generalization. Across five tasks of sequence labeling, we demonstrate that FISH-DIP can smoothly optimize the model in low resource settings offering upto 40{\%} performance improvements over full fine-tuning depending on target evaluation settings. Also, compared to in-context learning and other parameter-efficient fine-tuning approaches, FISH-DIP performs comparably or better, notably in extreme low-resource settings. The source code of FISH-DIP will be available at [this URL](https://github.com/psunlpgroup/FISH-DIP)

Unified Low-Resource Sequence Labeling by Sample-Aware Dynamic Sparse Finetuning

emnlp-main.433

2311.03748

Poster

https://aclanthology.org/2023.emnlp-main.434.bib

https://aclanthology.org/2023.emnlp-main.434/

@inproceedings{nowak-etal-2023-representational, title = "On the Representational Capacity of Recurrent Neural Language Models", author = "Nowak, Franz and Svete, Anej and Du, Li and Cotterell, Ryan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.434", doi = "10.18653/v1/2023.emnlp-main.434", pages = "7011--7034", abstract = "This work investigates the computational expressivity of language models (LMs) based on recurrent neural networks (RNNs). Siegelmann and Sontag (1992) famously showed that RNNs with rational weights and hidden states and unbounded computation time are Turing complete. However, LMs define weightings over strings in addition to just (unweighted) language membership and the analysis of the computational power of RNN LMs (RLMs) should reflect this. We extend the Turing completeness result to the probabilistic case, showing how a rationally weighted RLM with unbounded computation time can simulate any deterministic probabilistic Turing machine (PTM) with rationally weighted transitions. Since, in practice, RLMs work in real-time, processing a symbol at every time step, we treat the above result as an upper bound on the expressivity of RLMs. We also provide a lower bound by showing that under the restriction to real-time computation, such models can simulate deterministic real-time rational PTMs.", }

This work investigates the computational expressivity of language models (LMs) based on recurrent neural networks (RNNs). Siegelmann and Sontag (1992) famously showed that RNNs with rational weights and hidden states and unbounded computation time are Turing complete. However, LMs define weightings over strings in addition to just (unweighted) language membership and the analysis of the computational power of RNN LMs (RLMs) should reflect this. We extend the Turing completeness result to the probabilistic case, showing how a rationally weighted RLM with unbounded computation time can simulate any deterministic probabilistic Turing machine (PTM) with rationally weighted transitions. Since, in practice, RLMs work in real-time, processing a symbol at every time step, we treat the above result as an upper bound on the expressivity of RLMs. We also provide a lower bound by showing that under the restriction to real-time computation, such models can simulate deterministic real-time rational PTMs.

On the Representational Capacity of Recurrent Neural Language Models

emnlp-main.434

2310.12942

Poster

https://aclanthology.org/2023.emnlp-main.435.bib

https://aclanthology.org/2023.emnlp-main.435/

@inproceedings{stolfo-etal-2023-mechanistic, title = "A Mechanistic Interpretation of Arithmetic Reasoning in Language Models using Causal Mediation Analysis", author = "Stolfo, Alessandro and Belinkov, Yonatan and Sachan, Mrinmaya", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.435", doi = "10.18653/v1/2023.emnlp-main.435", pages = "7035--7052", abstract = "Mathematical reasoning in large language models (LMs) has garnered significant attention in recent work, but there is a limited understanding of how these models process and store information related to arithmetic tasks within their architecture. In order to improve our understanding of this aspect of language models, we present a mechanistic interpretation of Transformer-based LMs on arithmetic questions using a causal mediation analysis framework. By intervening on the activations of specific model components and measuring the resulting changes in predicted probabilities, we identify the subset of parameters responsible for specific predictions. This provides insights into how information related to arithmetic is processed by LMs. Our experimental results indicate that LMs process the input by transmitting the information relevant to the query from mid-sequence early layers to the final token using the attention mechanism. Then, this information is processed by a set of MLP modules, which generate result-related information that is incorporated into the residual stream. To assess the specificity of the observed activation dynamics, we compare the effects of different model components on arithmetic queries with other tasks, including number retrieval from prompts and factual knowledge questions.", }

Mathematical reasoning in large language models (LMs) has garnered significant attention in recent work, but there is a limited understanding of how these models process and store information related to arithmetic tasks within their architecture. In order to improve our understanding of this aspect of language models, we present a mechanistic interpretation of Transformer-based LMs on arithmetic questions using a causal mediation analysis framework. By intervening on the activations of specific model components and measuring the resulting changes in predicted probabilities, we identify the subset of parameters responsible for specific predictions. This provides insights into how information related to arithmetic is processed by LMs. Our experimental results indicate that LMs process the input by transmitting the information relevant to the query from mid-sequence early layers to the final token using the attention mechanism. Then, this information is processed by a set of MLP modules, which generate result-related information that is incorporated into the residual stream. To assess the specificity of the observed activation dynamics, we compare the effects of different model components on arithmetic queries with other tasks, including number retrieval from prompts and factual knowledge questions.

A Mechanistic Interpretation of Arithmetic Reasoning in Language Models using Causal Mediation Analysis

emnlp-main.435

2305.15054

https://huggingface.co/papers/2305.15054

Poster

https://aclanthology.org/2023.emnlp-main.436.bib

https://aclanthology.org/2023.emnlp-main.436/

@inproceedings{bhaskar-etal-2023-benchmarking, title = "Benchmarking and Improving Text-to-{SQL} Generation under Ambiguity", author = "Bhaskar, Adithya and Tomar, Tushar and Sathe, Ashutosh and Sarawagi, Sunita", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.436", doi = "10.18653/v1/2023.emnlp-main.436", pages = "7053--7074", abstract = "Research in Text-to-SQL conversion has been largely benchmarked against datasets where each text query corresponds to one correct SQL. However, natural language queries over real-life databases frequently involve significant ambiguity about the intended SQL due to overlapping schema names and multiple confusing relationship paths. To bridge this gap, we develop a novel benchmark called AmbiQT with over 3000 examples where each text is interpretable as two plausible SQLs due to lexical and/or structural ambiguity. When faced with ambiguity, an ideal top-$k$ decoder should generate all valid interpretations for possible disambiguation by the user. We evaluate several Text-to-SQL systems and decoding algorithms, including those employing state-of-the-art LLMs, and find them to be far from this ideal. The primary reason is that the prevalent beam search algorithm and its variants, treat SQL queries as a string and produce unhelpful token-level diversity in the top-$k$. We propose LogicalBeam, a new decoding algorithm that navigates the SQL logic space using a blend of plan-based template generation and constrained infilling. Counterfactually generated plans diversify templates while in-filling with a beam-search that branches solely on schema names provides value diversity. LogicalBeam is up to 2.5 times more effective than state-of-the-art models at generating all candidate SQLs in the top-$k$ ranked outputs. It also enhances the top-5 Exact and Execution Match Accuracies on SPIDER and Kaggle DBQA.", }

Research in Text-to-SQL conversion has been largely benchmarked against datasets where each text query corresponds to one correct SQL. However, natural language queries over real-life databases frequently involve significant ambiguity about the intended SQL due to overlapping schema names and multiple confusing relationship paths. To bridge this gap, we develop a novel benchmark called AmbiQT with over 3000 examples where each text is interpretable as two plausible SQLs due to lexical and/or structural ambiguity. When faced with ambiguity, an ideal top-$k$ decoder should generate all valid interpretations for possible disambiguation by the user. We evaluate several Text-to-SQL systems and decoding algorithms, including those employing state-of-the-art LLMs, and find them to be far from this ideal. The primary reason is that the prevalent beam search algorithm and its variants, treat SQL queries as a string and produce unhelpful token-level diversity in the top-$k$. We propose LogicalBeam, a new decoding algorithm that navigates the SQL logic space using a blend of plan-based template generation and constrained infilling. Counterfactually generated plans diversify templates while in-filling with a beam-search that branches solely on schema names provides value diversity. LogicalBeam is up to 2.5 times more effective than state-of-the-art models at generating all candidate SQLs in the top-$k$ ranked outputs. It also enhances the top-5 Exact and Execution Match Accuracies on SPIDER and Kaggle DBQA.

Benchmarking and Improving Text-to-SQL Generation under Ambiguity

emnlp-main.436

2310.13659

Poster

https://aclanthology.org/2023.emnlp-main.437.bib

https://aclanthology.org/2023.emnlp-main.437/

@inproceedings{zhang-etal-2023-non, title = "Non-autoregressive Text Editing with Copy-aware Latent Alignments", author = "Zhang, Yu and Zhang, Yue and Cui, Leyang and Fu, Guohong", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.437", doi = "10.18653/v1/2023.emnlp-main.437", pages = "7075--7085", abstract = "Recent work has witnessed a paradigm shift from Seq2Seq to Seq2Edit in the field of text editing, with the aim of addressing the slow autoregressive inference problem posed by the former. Despite promising results, Seq2Edit approaches still face several challenges such as inflexibility in generation and difficulty in generalizing to other languages. In this work, we propose a novel non-autoregressive text editing method to circumvent the above issues, by modeling the edit process with latent CTC alignments. We make a crucial extension to CTC by introducing the copy operation into the edit space, thus enabling more efficient management of textual overlap in editing. We conduct extensive experiments on GEC and sentence fusion tasks, showing that our proposed method significantly outperforms existing Seq2Edit models and achieves similar or even better results than Seq2Seq with over $4\times$ speedup. Moreover, it demonstrates good generalizability on German and Russian. In-depth analyses reveal the strengths of our method in terms of the robustness under various scenarios and generating fluent and flexible outputs.", }

Recent work has witnessed a paradigm shift from Seq2Seq to Seq2Edit in the field of text editing, with the aim of addressing the slow autoregressive inference problem posed by the former. Despite promising results, Seq2Edit approaches still face several challenges such as inflexibility in generation and difficulty in generalizing to other languages. In this work, we propose a novel non-autoregressive text editing method to circumvent the above issues, by modeling the edit process with latent CTC alignments. We make a crucial extension to CTC by introducing the copy operation into the edit space, thus enabling more efficient management of textual overlap in editing. We conduct extensive experiments on GEC and sentence fusion tasks, showing that our proposed method significantly outperforms existing Seq2Edit models and achieves similar or even better results than Seq2Seq with over $4\times$ speedup. Moreover, it demonstrates good generalizability on German and Russian. In-depth analyses reveal the strengths of our method in terms of the robustness under various scenarios and generating fluent and flexible outputs.

Non-autoregressive Text Editing with Copy-aware Latent Alignments

emnlp-main.437

2310.07821

https://huggingface.co/papers/2310.07821

Poster

https://aclanthology.org/2023.emnlp-main.438.bib

https://aclanthology.org/2023.emnlp-main.438/

@inproceedings{jalota-etal-2023-translating, title = "Translating away Translationese without Parallel Data", author = "Jalota, Rricha and Chowdhury, Koel and Espa{\~n}a-Bonet, Cristina and van Genabith, Josef", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.438", doi = "10.18653/v1/2023.emnlp-main.438", pages = "7086--7100", abstract = "Translated texts exhibit systematic linguistic differences compared to original texts in the same language, and these differences are referred to as translationese. Translationese has effects on various cross-lingual natural language processing tasks, potentially leading to biased results. In this paper, we explore a novel approach to reduce translationese in translated texts: translation-based style transfer. As there are no parallel human-translated and original data in the same language, we use a self-supervised approach that can learn from comparable (rather than parallel) mono-lingual original and translated data. However, even this self-supervised approach requires some parallel data for validation. We show how we can eliminate the need for parallel validation data by combining the self-supervised loss with an unsupervised loss. This unsupervised loss leverages the original language model loss over the style-transferred output and a semantic similarity loss between the input and style-transferred output. We evaluate our approach in terms of original vs. translationese binary classification in addition to measuring content preservation and target-style fluency. The results show that our approach is able to reduce translationese classifier accuracy to a level of a random classifier after style transfer while adequately preserving the content and fluency in the target original style.", }

Translated texts exhibit systematic linguistic differences compared to original texts in the same language, and these differences are referred to as translationese. Translationese has effects on various cross-lingual natural language processing tasks, potentially leading to biased results. In this paper, we explore a novel approach to reduce translationese in translated texts: translation-based style transfer. As there are no parallel human-translated and original data in the same language, we use a self-supervised approach that can learn from comparable (rather than parallel) mono-lingual original and translated data. However, even this self-supervised approach requires some parallel data for validation. We show how we can eliminate the need for parallel validation data by combining the self-supervised loss with an unsupervised loss. This unsupervised loss leverages the original language model loss over the style-transferred output and a semantic similarity loss between the input and style-transferred output. We evaluate our approach in terms of original vs. translationese binary classification in addition to measuring content preservation and target-style fluency. The results show that our approach is able to reduce translationese classifier accuracy to a level of a random classifier after style transfer while adequately preserving the content and fluency in the target original style.

Translating away Translationese without Parallel Data

emnlp-main.438

2310.18830

Oral

https://aclanthology.org/2023.emnlp-main.439.bib

https://aclanthology.org/2023.emnlp-main.439/

@inproceedings{yu-etal-2023-prompt, title = "Prompt-Based {M}onte-{C}arlo Tree Search for Goal-oriented Dialogue Policy Planning", author = "Yu, Xiao and Chen, Maximillian and Yu, Zhou", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.439", doi = "10.18653/v1/2023.emnlp-main.439", pages = "7101--7125", abstract = "Planning for goal-oriented dialogue often requires simulating future dialogue interactions and estimating task progress. Many approaches thus consider training neural networks to perform look-ahead search algorithms such as A* search and Monte Carlo Tree Search (MCTS). However, this training often require abundant annotated data, which creates challenges when faced with noisy annotations or low-resource settings. We introduce GDP-Zero, an approach using Open-Loop MCTS to perform goal-oriented dialogue policy planning without any model training. GDP-Zero prompts a large language model to act as a policy prior, value function, user simulator, and system model during the tree search. We evaluate GDP-Zero on the goal-oriented task PersuasionForGood, and find that its responses are preferred over ChatGPT up to 59.32{\%} of the time, and are rated more persuasive than ChatGPT during interactive evaluations.", }

Planning for goal-oriented dialogue often requires simulating future dialogue interactions and estimating task progress. Many approaches thus consider training neural networks to perform look-ahead search algorithms such as A* search and Monte Carlo Tree Search (MCTS). However, this training often require abundant annotated data, which creates challenges when faced with noisy annotations or low-resource settings. We introduce GDP-Zero, an approach using Open-Loop MCTS to perform goal-oriented dialogue policy planning without any model training. GDP-Zero prompts a large language model to act as a policy prior, value function, user simulator, and system model during the tree search. We evaluate GDP-Zero on the goal-oriented task PersuasionForGood, and find that its responses are preferred over ChatGPT up to 59.32{\%} of the time, and are rated more persuasive than ChatGPT during interactive evaluations.

Prompt-Based Monte-Carlo Tree Search for Goal-oriented Dialogue Policy Planning

emnlp-main.439

2305.13660

https://huggingface.co/papers/2305.13660

Oral

https://aclanthology.org/2023.emnlp-main.440.bib

https://aclanthology.org/2023.emnlp-main.440/

@inproceedings{liang-etal-2023-unimath, title = "{U}ni{M}ath: A Foundational and Multimodal Mathematical Reasoner", author = "Liang, Zhenwen and Yang, Tianyu and Zhang, Jipeng and Zhang, Xiangliang", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.440", doi = "10.18653/v1/2023.emnlp-main.440", pages = "7126--7133", abstract = "While significant progress has been made in natural language processing (NLP), existing methods exhibit limitations in effectively interpreting and processing diverse mathematical modalities. Therefore, we introduce UniMath, a versatile and unified system designed for multimodal mathematical reasoning tasks. Tackling complex problem-solving in arithmetic, geometry, and table-based math, UniMath utilizes a fine-tuned T5 model augmented with a variational autoencoder (VAE)-based image tokenizer. By jointly training and evaluating the model on three diverse datasets - SVAMP, GeoQA, and TableMWP, UniMath achieves state-of-the-art performance. The model{'}s generalization ability is further demonstrated via fine-tuning on two additional datasets, MathQA and Geo-Proving. Through comprehensive evaluations, we showcase that joint training across diverse math tasks improves overall model performance and enhances its ability to generalize across different mathematical reasoning tasks. This pioneering approach provides a blueprint and inspires further efforts on unified mathematical reasoning with deep learning systems.", }

While significant progress has been made in natural language processing (NLP), existing methods exhibit limitations in effectively interpreting and processing diverse mathematical modalities. Therefore, we introduce UniMath, a versatile and unified system designed for multimodal mathematical reasoning tasks. Tackling complex problem-solving in arithmetic, geometry, and table-based math, UniMath utilizes a fine-tuned T5 model augmented with a variational autoencoder (VAE)-based image tokenizer. By jointly training and evaluating the model on three diverse datasets - SVAMP, GeoQA, and TableMWP, UniMath achieves state-of-the-art performance. The model{'}s generalization ability is further demonstrated via fine-tuning on two additional datasets, MathQA and Geo-Proving. Through comprehensive evaluations, we showcase that joint training across diverse math tasks improves overall model performance and enhances its ability to generalize across different mathematical reasoning tasks. This pioneering approach provides a blueprint and inspires further efforts on unified mathematical reasoning with deep learning systems.

UniMath: A Foundational and Multimodal Mathematical Reasoner

emnlp-main.440

Oral

https://aclanthology.org/2023.emnlp-main.441.bib

https://aclanthology.org/2023.emnlp-main.441/

@inproceedings{ma-etal-2023-caseencoder, title = "{C}ase{E}ncoder: A Knowledge-enhanced Pre-trained Model for Legal Case Encoding", author = "Ma, Yixiao and Wu, Yueyue and Su, Weihang and Ai, Qingyao and Liu, Yiqun", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.441", doi = "10.18653/v1/2023.emnlp-main.441", pages = "7134--7143", abstract = "Legal case retrieval is a critical process for modern legal information systems. While recent studies have utilized pre-trained language models (PLMs) based on the general domain self-supervised pre-training paradigm to build models for legal case retrieval, there are limitations in using general domain PLMs as backbones. Specifically, these models may not fully capture the underlying legal features in legal case documents. To address this issue, we propose CaseEncoder, a legal document encoder that leverages fine-grained legal knowledge in both the data sampling and pre-training phases. In the data sampling phase, we enhance the quality of the training data by utilizing fine-grained law article information to guide the selection of positive and negative examples. In the pre-training phase, we design legal-specific pre-training tasks that align with the judging criteria of relevant legal cases. Based on these tasks, we introduce an innovative loss function called Biased Circle Loss to enhance the model{'}s ability to recognize case relevance in fine grains. Experimental results on multiple benchmarks demonstrate that CaseEncoder significantly outperforms both existing general pre-training models and legal-specific pre-training models in zero-shot legal case retrieval. The source code of CaseEncoder can be found at https://github.com/Anonymous-EMNLP2023/CaseEncoder.", }

Legal case retrieval is a critical process for modern legal information systems. While recent studies have utilized pre-trained language models (PLMs) based on the general domain self-supervised pre-training paradigm to build models for legal case retrieval, there are limitations in using general domain PLMs as backbones. Specifically, these models may not fully capture the underlying legal features in legal case documents. To address this issue, we propose CaseEncoder, a legal document encoder that leverages fine-grained legal knowledge in both the data sampling and pre-training phases. In the data sampling phase, we enhance the quality of the training data by utilizing fine-grained law article information to guide the selection of positive and negative examples. In the pre-training phase, we design legal-specific pre-training tasks that align with the judging criteria of relevant legal cases. Based on these tasks, we introduce an innovative loss function called Biased Circle Loss to enhance the model{'}s ability to recognize case relevance in fine grains. Experimental results on multiple benchmarks demonstrate that CaseEncoder significantly outperforms both existing general pre-training models and legal-specific pre-training models in zero-shot legal case retrieval. The source code of CaseEncoder can be found at https://github.com/Anonymous-EMNLP2023/CaseEncoder.

CaseEncoder: A Knowledge-enhanced Pre-trained Model for Legal Case Encoding

emnlp-main.441

2305.05393

Poster

https://aclanthology.org/2023.emnlp-main.442.bib

https://aclanthology.org/2023.emnlp-main.442/

@inproceedings{watson-etal-2023-hiddentables, title = "{H}idden{T}ables and {P}y{QT}ax: A Cooperative Game and Dataset For {T}able{QA} to Ensure Scale and Data Privacy Across a Myriad of Taxonomies", author = "Watson, William and Cho, Nicole and Balch, Tucker and Veloso, Manuela", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.442", doi = "10.18653/v1/2023.emnlp-main.442", pages = "7144--7159", abstract = "A myriad of different Large Language Models (LLMs) face a common challenge in contextually analyzing table question-answering tasks. These challenges are engendered from (1) finite context windows for large tables, (2) multi-faceted discrepancies amongst tokenization patterns against cell boundaries, and (3) various limitations stemming from data confidentiality in the process of using external models such as gpt-35-turbo. We propose a cooperative game dubbed {``}HiddenTables{''} as a potential resolution to this challenge. In essence, {``}HiddenTables{''} is played between the code-generating LLM {``}Solver{''} and the {``}Oracle{''} which evaluates the ability of the LLM agents to solve TableQA tasks. This game is based on natural language schemas and importantly, ensures the security of the underlying data. We provide evidential experiments on a diverse set of tables that demonstrate an LLM{'}s collective inability to generalize and perform on complex queries, handle compositional dependencies, and align natural language to programmatic commands when concrete table schemas are provided. Unlike encoder-based models, we have pushed the boundaries of {``}HiddenTables{''} to not be limited by the number of rows - therefore we exhibit improved efficiency in prompt and completion tokens. Our infrastructure has spawned a new dataset {``}PyQTax{''} that spans across 116,671 question-table-answer triplets and provides additional fine-grained breakdowns and labels for varying question taxonomies. Therefore, in tandem with our academic contributions regarding LLMs{'} deficiency in TableQA tasks, {``}HiddenTables{''} is a tactile manifestation of how LLMs can interact with massive datasets while ensuring data security and minimizing generation costs.", }

A myriad of different Large Language Models (LLMs) face a common challenge in contextually analyzing table question-answering tasks. These challenges are engendered from (1) finite context windows for large tables, (2) multi-faceted discrepancies amongst tokenization patterns against cell boundaries, and (3) various limitations stemming from data confidentiality in the process of using external models such as gpt-35-turbo. We propose a cooperative game dubbed {``}HiddenTables{''} as a potential resolution to this challenge. In essence, {``}HiddenTables{''} is played between the code-generating LLM {``}Solver{''} and the {``}Oracle{''} which evaluates the ability of the LLM agents to solve TableQA tasks. This game is based on natural language schemas and importantly, ensures the security of the underlying data. We provide evidential experiments on a diverse set of tables that demonstrate an LLM{'}s collective inability to generalize and perform on complex queries, handle compositional dependencies, and align natural language to programmatic commands when concrete table schemas are provided. Unlike encoder-based models, we have pushed the boundaries of {``}HiddenTables{''} to not be limited by the number of rows - therefore we exhibit improved efficiency in prompt and completion tokens. Our infrastructure has spawned a new dataset {``}PyQTax{''} that spans across 116,671 question-table-answer triplets and provides additional fine-grained breakdowns and labels for varying question taxonomies. Therefore, in tandem with our academic contributions regarding LLMs{'} deficiency in TableQA tasks, {``}HiddenTables{''} is a tactile manifestation of how LLMs can interact with massive datasets while ensuring data security and minimizing generation costs.

HiddenTables and PyQTax: A Cooperative Game and Dataset For TableQA to Ensure Scale and Data Privacy Across a Myriad of Taxonomies

emnlp-main.442

2406.10803

https://huggingface.co/papers/2406.10803

Poster

https://aclanthology.org/2023.emnlp-main.443.bib

https://aclanthology.org/2023.emnlp-main.443/

@inproceedings{zhao-etal-2023-causal, title = "Causal Document-Grounded Dialogue Pre-training", author = "Zhao, Yingxiu and Yu, Bowen and Li, Bowen and Yu, Haiyang and Li, Jinyang and Wang, Chao and Huang, Fei and Li, Yongbin and Zhang, Nevin", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.443", doi = "10.18653/v1/2023.emnlp-main.443", pages = "7160--7174", abstract = "The goal of document-grounded dialogue (DocGD) is to generate a response by anchoring the evidence in a supporting document in accordance with the dialogue context. This entails four causally interconnected variables. While task-specific pre-training has significantly enhanced performances on numerous downstream tasks, existing DocGD methods still rely on general pre-trained language models without a specifically tailored pre-training approach that explicitly captures the causal relationships. To address this, we present the first causally-complete dataset construction strategy for developing million-scale DocGD pre-training corpora. Additionally, we propose a causally-perturbed pre-training strategy to better capture causality by introducing perturbations on the variables and optimizing the overall causal effect. Experiments conducted on three benchmark datasets demonstrate that our causal pre-training yields substantial and consistent improvements in fully-supervised, low-resource, few-shot, and zero-shot settings.", }

The goal of document-grounded dialogue (DocGD) is to generate a response by anchoring the evidence in a supporting document in accordance with the dialogue context. This entails four causally interconnected variables. While task-specific pre-training has significantly enhanced performances on numerous downstream tasks, existing DocGD methods still rely on general pre-trained language models without a specifically tailored pre-training approach that explicitly captures the causal relationships. To address this, we present the first causally-complete dataset construction strategy for developing million-scale DocGD pre-training corpora. Additionally, we propose a causally-perturbed pre-training strategy to better capture causality by introducing perturbations on the variables and optimizing the overall causal effect. Experiments conducted on three benchmark datasets demonstrate that our causal pre-training yields substantial and consistent improvements in fully-supervised, low-resource, few-shot, and zero-shot settings.

Causal Document-Grounded Dialogue Pre-training

emnlp-main.443

2305.10927

Poster

https://aclanthology.org/2023.emnlp-main.444.bib

https://aclanthology.org/2023.emnlp-main.444/

@inproceedings{prabhu-etal-2023-accented, title = "Accented Speech Recognition With Accent-specific Codebooks", author = "Prabhu, Darshan and Jyothi, Preethi and Ganapathy, Sriram and Unni, Vinit", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.444", doi = "10.18653/v1/2023.emnlp-main.444", pages = "7175--7188", abstract = "Speech accents pose a significant challenge to state-of-the-art automatic speech recognition (ASR) systems. Degradation in performance across underrepresented accents is a severe deterrent to the inclusive adoption of ASR. In this work, we propose a novel accent adaptation approach for end-to-end ASR systems using cross-attention with a trainable set of codebooks. These learnable codebooks capture accent-specific information and are integrated within the ASR encoder layers. The model is trained on accented English speech, while the test data also contained accents which were not seen during training. On the Mozilla Common Voice multi-accented dataset, we show that our proposed approach yields significant performance gains not only on the seen English accents (up to 37{\%} relative improvement in word error rate) but also on the unseen accents (up to 5{\%} relative improvement in WER). Further, we illustrate benefits for a zero-shot transfer setup on the L2Artic dataset. We also compare the performance with other approaches based on accent adversarial training.", }

Speech accents pose a significant challenge to state-of-the-art automatic speech recognition (ASR) systems. Degradation in performance across underrepresented accents is a severe deterrent to the inclusive adoption of ASR. In this work, we propose a novel accent adaptation approach for end-to-end ASR systems using cross-attention with a trainable set of codebooks. These learnable codebooks capture accent-specific information and are integrated within the ASR encoder layers. The model is trained on accented English speech, while the test data also contained accents which were not seen during training. On the Mozilla Common Voice multi-accented dataset, we show that our proposed approach yields significant performance gains not only on the seen English accents (up to 37{\%} relative improvement in word error rate) but also on the unseen accents (up to 5{\%} relative improvement in WER). Further, we illustrate benefits for a zero-shot transfer setup on the L2Artic dataset. We also compare the performance with other approaches based on accent adversarial training.

Accented Speech Recognition With Accent-specific Codebooks

emnlp-main.444

2310.15970

Poster

https://aclanthology.org/2023.emnlp-main.445.bib

https://aclanthology.org/2023.emnlp-main.445/

@inproceedings{radevski-etal-2023-linking, title = "Linking Surface Facts to Large-Scale Knowledge Graphs", author = "Radevski, Gorjan and Gashteovski, Kiril and Hung, Chia-Chien and Lawrence, Carolin and Glava{\v{s}}, Goran", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.445", doi = "10.18653/v1/2023.emnlp-main.445", pages = "7189--7207", abstract = "Open Information Extraction (OIE) methods extract facts from natural language text in the form of ({``}subject{''}; {``}relation{''}; {``}object{''}) triples. These facts are, however, merely surface forms, the ambiguity of which impedes their downstream usage; e.g., the surface phrase {``}Michael Jordan{''} may refer to either the former basketball player or the university professor. Knowledge Graphs (KGs), on the other hand, contain facts in a canonical (i.e., unambiguous) form, but their coverage is limited by a static schema (i.e., a fixed set of entities and predicates). To bridge this gap, we need the best of both worlds: (i) high coverage of free-text OIEs, and (ii) semantic precision (i.e., monosemy) of KGs. In order to achieve this goal, we propose a new benchmark with novel evaluation protocols that can, for example, measure fact linking performance on a granular triple slot level, while also measuring if a system has the ability to recognize that a surface form has no match in the existing KG. Our extensive evaluation of several baselines show that detection of out-of-KG entities and predicates is more difficult than accurate linking to existing ones, thus calling for more research efforts on this difficult task. We publicly release all resources (data, benchmark and code) on https://github.com/nec-research/fact-linking.", }

Open Information Extraction (OIE) methods extract facts from natural language text in the form of ({``}subject{''}; {``}relation{''}; {``}object{''}) triples. These facts are, however, merely surface forms, the ambiguity of which impedes their downstream usage; e.g., the surface phrase {``}Michael Jordan{''} may refer to either the former basketball player or the university professor. Knowledge Graphs (KGs), on the other hand, contain facts in a canonical (i.e., unambiguous) form, but their coverage is limited by a static schema (i.e., a fixed set of entities and predicates). To bridge this gap, we need the best of both worlds: (i) high coverage of free-text OIEs, and (ii) semantic precision (i.e., monosemy) of KGs. In order to achieve this goal, we propose a new benchmark with novel evaluation protocols that can, for example, measure fact linking performance on a granular triple slot level, while also measuring if a system has the ability to recognize that a surface form has no match in the existing KG. Our extensive evaluation of several baselines show that detection of out-of-KG entities and predicates is more difficult than accurate linking to existing ones, thus calling for more research efforts on this difficult task. We publicly release all resources (data, benchmark and code) on https://github.com/nec-research/fact-linking.

Linking Surface Facts to Large-Scale Knowledge Graphs

emnlp-main.445

2310.14909

Poster

https://aclanthology.org/2023.emnlp-main.446.bib

https://aclanthology.org/2023.emnlp-main.446/

@inproceedings{zhang-etal-2023-sentiment-analysis, title = "Sentiment Analysis on Streaming User Reviews via Dual-Channel Dynamic Graph Neural Network", author = "Zhang, Xin and Zhang, Linhai and Zhou, Deyu", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.446", doi = "10.18653/v1/2023.emnlp-main.446", pages = "7208--7220", abstract = "Sentiment analysis on user reviews has achieved great success thanks to the rapid growth of deep learning techniques. The large number of online streaming reviews also provides the opportunity to model temporal dynamics for users and products on the timeline. However, existing methods model users and products in the real world based on a static assumption and neglect their time-varying characteristics. In this paper, we present DC-DGNN, a dual-channel framework based on a dynamic graph neural network (DGNN) that models temporal user and product dynamics for sentiment analysis. Specifically, a dual-channel text encoder is employed to extract current local and global contexts from review documents for users and products. Moreover, user review streams are integrated into the dynamic graph neural network by treating users and products as nodes and reviews as new edges. Node representations are dynamically updated along with the evolution of the dynamic graph and used for the final score prediction. Experimental results on five real-world datasets demonstrate the superiority of the proposed method.", }

Sentiment analysis on user reviews has achieved great success thanks to the rapid growth of deep learning techniques. The large number of online streaming reviews also provides the opportunity to model temporal dynamics for users and products on the timeline. However, existing methods model users and products in the real world based on a static assumption and neglect their time-varying characteristics. In this paper, we present DC-DGNN, a dual-channel framework based on a dynamic graph neural network (DGNN) that models temporal user and product dynamics for sentiment analysis. Specifically, a dual-channel text encoder is employed to extract current local and global contexts from review documents for users and products. Moreover, user review streams are integrated into the dynamic graph neural network by treating users and products as nodes and reviews as new edges. Node representations are dynamically updated along with the evolution of the dynamic graph and used for the final score prediction. Experimental results on five real-world datasets demonstrate the superiority of the proposed method.

Sentiment Analysis on Streaming User Reviews via Dual-Channel Dynamic Graph Neural Network

emnlp-main.446

Poster

https://aclanthology.org/2023.emnlp-main.447.bib

https://aclanthology.org/2023.emnlp-main.447/

@inproceedings{de-vries-etal-2023-dumb, title = "{DUMB}: A Benchmark for Smart Evaluation of {D}utch Models", author = "de Vries, Wietse and Wieling, Martijn and Nissim, Malvina", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.447", doi = "10.18653/v1/2023.emnlp-main.447", pages = "7221--7241", abstract = "We introduce the Dutch Model Benchmark: DUMB. The benchmark includes a diverse set of datasets for low-, medium- and high-resource tasks. The total set of nine tasks includes four tasks that were previously not available in Dutch. Instead of relying on a mean score across tasks, we propose Relative Error Reduction (RER), which compares the DUMB performance of language models to a strong baseline which can be referred to in the future even when assessing different sets of language models. Through a comparison of 14 pre-trained language models (mono- and multi-lingual, of varying sizes), we assess the internal consistency of the benchmark tasks, as well as the factors that likely enable high performance. Our results indicate that current Dutch monolingual models under-perform and suggest training larger Dutch models with other architectures and pre-training objectives. At present, the highest performance is achieved by DeBERTaV3 (large), XLM-R (large) and mDeBERTaV3 (base). In addition to highlighting best strategies for training larger Dutch models, DUMB will foster further research on Dutch. A public leaderboard is available at https://dumbench.nl.", }

We introduce the Dutch Model Benchmark: DUMB. The benchmark includes a diverse set of datasets for low-, medium- and high-resource tasks. The total set of nine tasks includes four tasks that were previously not available in Dutch. Instead of relying on a mean score across tasks, we propose Relative Error Reduction (RER), which compares the DUMB performance of language models to a strong baseline which can be referred to in the future even when assessing different sets of language models. Through a comparison of 14 pre-trained language models (mono- and multi-lingual, of varying sizes), we assess the internal consistency of the benchmark tasks, as well as the factors that likely enable high performance. Our results indicate that current Dutch monolingual models under-perform and suggest training larger Dutch models with other architectures and pre-training objectives. At present, the highest performance is achieved by DeBERTaV3 (large), XLM-R (large) and mDeBERTaV3 (base). In addition to highlighting best strategies for training larger Dutch models, DUMB will foster further research on Dutch. A public leaderboard is available at https://dumbench.nl.

DUMB: A Benchmark for Smart Evaluation of Dutch Models

emnlp-main.447

2305.13026

Poster

https://aclanthology.org/2023.emnlp-main.448.bib

https://aclanthology.org/2023.emnlp-main.448/

@inproceedings{shi-etal-2023-osscse, title = "{O}ss{CSE}: Overcoming Surface Structure Bias in Contrastive Learning for Unsupervised Sentence Embedding", author = "Shi, Zhan and Wang, Guoyin and Bai, Ke and Li, Jiwei and Li, Xiang and Cui, Qingjun and Zeng, Belinda and Chilimbi, Trishul and Zhu, Xiaodan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.448", doi = "10.18653/v1/2023.emnlp-main.448", pages = "7242--7254", abstract = "Contrastive learning has been demonstrated effective in unsupervised sentence representation learning. Given one sentence, positive pairs are obtained by passing the sentence to the encoder twice using the different dropout masks, and negative pairs are obtained by taking another sentence in the same mini-batch. However, the method suffers from the surface structure bias, i.e., sentences with similar surface structures will be regarded as close in semantics while sentences with dissimilar surface structures will be viewed as distinct in semantics. This leads to the result that paraphrasing a sentence that is dissimilar in surface structure will receive a lower semantic similarity score than inserting a negative word into the sentence. In this paper, we first verify the bias by collecting a sentence transformation testset. Then we systematically probe the existing models by proposing novel splits based on benchmark datasets in accordance with semantic and surface structure similarity. We tackle the bias in two aspects: balancing the learning target by augmenting with data that counters the bias, and meanwhile preserving word semantics by leveraging recall loss to prevent catastrophic forgetting. We evaluate our model on standard semantic textual similarity (STS) tasks using different pre-trained backbones and achieve state-of-the-art averaged performance across the STS benchmarks. Particularly, our models that are fine-tuned with $RoBERTa_{base}$ and $RoBERTa_{large}$ achieve significantly better performance on most benchmark datasets.", }

Contrastive learning has been demonstrated effective in unsupervised sentence representation learning. Given one sentence, positive pairs are obtained by passing the sentence to the encoder twice using the different dropout masks, and negative pairs are obtained by taking another sentence in the same mini-batch. However, the method suffers from the surface structure bias, i.e., sentences with similar surface structures will be regarded as close in semantics while sentences with dissimilar surface structures will be viewed as distinct in semantics. This leads to the result that paraphrasing a sentence that is dissimilar in surface structure will receive a lower semantic similarity score than inserting a negative word into the sentence. In this paper, we first verify the bias by collecting a sentence transformation testset. Then we systematically probe the existing models by proposing novel splits based on benchmark datasets in accordance with semantic and surface structure similarity. We tackle the bias in two aspects: balancing the learning target by augmenting with data that counters the bias, and meanwhile preserving word semantics by leveraging recall loss to prevent catastrophic forgetting. We evaluate our model on standard semantic textual similarity (STS) tasks using different pre-trained backbones and achieve state-of-the-art averaged performance across the STS benchmarks. Particularly, our models that are fine-tuned with $RoBERTa_{base}$ and $RoBERTa_{large}$ achieve significantly better performance on most benchmark datasets.

OssCSE: Overcoming Surface Structure Bias in Contrastive Learning for Unsupervised Sentence Embedding

emnlp-main.448

Poster

https://aclanthology.org/2023.emnlp-main.449.bib

https://aclanthology.org/2023.emnlp-main.449/

@inproceedings{zuluaga-gomez-etal-2023-end, title = "End-to-End Single-Channel Speaker-Turn Aware Conversational Speech Translation", author = "Zuluaga-Gomez, Juan Pablo and Huang, Zhaocheng and Niu, Xing and Paturi, Rohit and Srinivasan, Sundararajan and Mathur, Prashant and Thompson, Brian and Federico, Marcello", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.449", doi = "10.18653/v1/2023.emnlp-main.449", pages = "7255--7274", abstract = "Conventional speech-to-text translation (ST) systems are trained on single-speaker utterances, and they may not generalize to real-life scenarios where the audio contains conversations by multiple speakers. In this paper, we tackle single-channel multi-speaker conversational ST with an end-to-end and multi-task training model, named Speaker-Turn Aware Conversational Speech Translation, that combines automatic speech recognition, speech translation and speaker turn detection using special tokens in a serialized labeling format. We run experiments on the Fisher-CALLHOME corpus, which we adapted by merging the two single-speaker channels into one multi-speaker channel, thus representing the more realistic and challenging scenario with multi-speaker turns and cross-talk. Experimental results across single- and multi-speaker conditions and against conventional ST systems, show that our model outperforms the reference systems on the multi-speaker condition, while attaining comparable performance on the single-speaker condition. We release scripts for data processing and model training.", }

Conventional speech-to-text translation (ST) systems are trained on single-speaker utterances, and they may not generalize to real-life scenarios where the audio contains conversations by multiple speakers. In this paper, we tackle single-channel multi-speaker conversational ST with an end-to-end and multi-task training model, named Speaker-Turn Aware Conversational Speech Translation, that combines automatic speech recognition, speech translation and speaker turn detection using special tokens in a serialized labeling format. We run experiments on the Fisher-CALLHOME corpus, which we adapted by merging the two single-speaker channels into one multi-speaker channel, thus representing the more realistic and challenging scenario with multi-speaker turns and cross-talk. Experimental results across single- and multi-speaker conditions and against conventional ST systems, show that our model outperforms the reference systems on the multi-speaker condition, while attaining comparable performance on the single-speaker condition. We release scripts for data processing and model training.

End-to-End Single-Channel Speaker-Turn Aware Conversational Speech Translation

emnlp-main.449

2311.00697

Poster

https://aclanthology.org/2023.emnlp-main.450.bib

https://aclanthology.org/2023.emnlp-main.450/

@inproceedings{yu-etal-2023-fine, title = "A Fine-Grained Taxonomy of Replies to Hate Speech", author = "Yu, Xinchen and Zhao, Ashley and Blanco, Eduardo and Hong, Lingzi", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.450", doi = "10.18653/v1/2023.emnlp-main.450", pages = "7275--7289", abstract = "Countering rather than censoring hate speech has emerged as a promising strategy to address hatred. There are many types of counterspeech in user-generated content: addressing the hateful content or its author, generic requests, well-reasoned counter arguments, insults, etc. The effectiveness of counterspeech, which we define as subsequent incivility, depends on these types. In this paper, we present a theoretically grounded taxonomy of replies to hate speech and a new corpus. We work with real, user-generated hate speech and all the replies it elicits rather than replies generated by a third party. Our analyses provide insights into the content real users reply with as well as which replies are empirically most effective. We also experiment with models to characterize the replies to hate speech, thereby opening the door to estimating whether a reply to hate speech will result in further incivility.", }

Countering rather than censoring hate speech has emerged as a promising strategy to address hatred. There are many types of counterspeech in user-generated content: addressing the hateful content or its author, generic requests, well-reasoned counter arguments, insults, etc. The effectiveness of counterspeech, which we define as subsequent incivility, depends on these types. In this paper, we present a theoretically grounded taxonomy of replies to hate speech and a new corpus. We work with real, user-generated hate speech and all the replies it elicits rather than replies generated by a third party. Our analyses provide insights into the content real users reply with as well as which replies are empirically most effective. We also experiment with models to characterize the replies to hate speech, thereby opening the door to estimating whether a reply to hate speech will result in further incivility.

A Fine-Grained Taxonomy of Replies to Hate Speech

emnlp-main.450

Poster

https://aclanthology.org/2023.emnlp-main.451.bib

https://aclanthology.org/2023.emnlp-main.451/

@inproceedings{zou-caragea-2023-jointmatch, title = "{J}oint{M}atch: A Unified Approach for Diverse and Collaborative Pseudo-Labeling to Semi-Supervised Text Classification", author = "Zou, Henry and Caragea, Cornelia", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.451", doi = "10.18653/v1/2023.emnlp-main.451", pages = "7290--7301", abstract = "Semi-supervised text classification (SSTC) has gained increasing attention due to its ability to leverage unlabeled data. However, existing approaches based on pseudo-labeling suffer from the issues of pseudo-label bias and error accumulation. In this paper, we propose JointMatch, a holistic approach for SSTC that addresses these challenges by unifying ideas from recent semi-supervised learning and the task of learning with noise. JointMatch adaptively adjusts classwise thresholds based on the learning status of different classes to mitigate model bias towards current easy classes. Additionally, JointMatch alleviates error accumulation by utilizing two differently initialized networks to teach each other in a cross-labeling manner. To maintain divergence between the two networks for mutual learning, we introduce a strategy that weighs more disagreement data while also allowing the utilization of high-quality agreement data for training. Experimental results on benchmark datasets demonstrate the superior performance of JointMatch, achieving a significant 5.13{\%} improvement on average. Notably, JointMatch delivers impressive results even in the extremely-scarce-label setting, obtaining 86{\%} accuracy on AG News with only 5 labels per class. We make our code available at https://github.com/HenryPengZou/JointMatch.", }

Semi-supervised text classification (SSTC) has gained increasing attention due to its ability to leverage unlabeled data. However, existing approaches based on pseudo-labeling suffer from the issues of pseudo-label bias and error accumulation. In this paper, we propose JointMatch, a holistic approach for SSTC that addresses these challenges by unifying ideas from recent semi-supervised learning and the task of learning with noise. JointMatch adaptively adjusts classwise thresholds based on the learning status of different classes to mitigate model bias towards current easy classes. Additionally, JointMatch alleviates error accumulation by utilizing two differently initialized networks to teach each other in a cross-labeling manner. To maintain divergence between the two networks for mutual learning, we introduce a strategy that weighs more disagreement data while also allowing the utilization of high-quality agreement data for training. Experimental results on benchmark datasets demonstrate the superior performance of JointMatch, achieving a significant 5.13{\%} improvement on average. Notably, JointMatch delivers impressive results even in the extremely-scarce-label setting, obtaining 86{\%} accuracy on AG News with only 5 labels per class. We make our code available at https://github.com/HenryPengZou/JointMatch.

JointMatch: A Unified Approach for Diverse and Collaborative Pseudo-Labeling to Semi-Supervised Text Classification

emnlp-main.451

2310.14583

Poster

https://aclanthology.org/2023.emnlp-main.452.bib

https://aclanthology.org/2023.emnlp-main.452/

@inproceedings{mireshghallah-etal-2023-simple, title = "Simple Temporal Adaptation to Changing Label Sets: Hashtag Prediction via Dense {KNN}", author = "Mireshghallah, Niloofar and Vogler, Nikolai and He, Junxian and Florez, Omar and El-Kishky, Ahmed and Berg-Kirkpatrick, Taylor", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.452", doi = "10.18653/v1/2023.emnlp-main.452", pages = "7302--7311", abstract = "User-generated social media data is constantly changing as new trends influence online discussion and personal information is deleted due to privacy concerns. However, traditional NLP models rely on fixed training datasets, which means they are unable to adapt to temporal change{---}both test distribution shift and deleted training data{---}without frequent, costly re-training. In this paper, we study temporal adaptation through the task of longitudinal hashtag prediction and propose a non-parametric dense retrieval technique, which does not require re-training, as a simple but effective solution. In experiments on a newly collected, publicly available, year-long Twitter dataset exhibiting temporal distribution shift, our method improves by 64{\%} over the best static parametric baseline while avoiding costly gradient-based re-training. Our approach is also particularly well-suited to dynamically deleted user data in line with data privacy laws, with negligible computational cost/performance loss.", }

User-generated social media data is constantly changing as new trends influence online discussion and personal information is deleted due to privacy concerns. However, traditional NLP models rely on fixed training datasets, which means they are unable to adapt to temporal change{---}both test distribution shift and deleted training data{---}without frequent, costly re-training. In this paper, we study temporal adaptation through the task of longitudinal hashtag prediction and propose a non-parametric dense retrieval technique, which does not require re-training, as a simple but effective solution. In experiments on a newly collected, publicly available, year-long Twitter dataset exhibiting temporal distribution shift, our method improves by 64{\%} over the best static parametric baseline while avoiding costly gradient-based re-training. Our approach is also particularly well-suited to dynamically deleted user data in line with data privacy laws, with negligible computational cost/performance loss.

Simple Temporal Adaptation to Changing Label Sets: Hashtag Prediction via Dense KNN

emnlp-main.452

Poster

https://aclanthology.org/2023.emnlp-main.453.bib

https://aclanthology.org/2023.emnlp-main.453/

@inproceedings{chang-etal-2023-speak, title = "Speak, Memory: An Archaeology of Books Known to {C}hat{GPT}/{GPT}-4", author = "Chang, Kent and Cramer, Mackenzie and Soni, Sandeep and Bamman, David", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.453", doi = "10.18653/v1/2023.emnlp-main.453", pages = "7312--7327", abstract = "In this work, we carry out a data archaeology to infer books that are known to ChatGPT and GPT-4 using a name cloze membership inference query. We find that OpenAI models have memorized a wide collection of copyrighted materials, and that the degree of memorization is tied to the frequency with which passages of those books appear on the web. The ability of these models to memorize an unknown set of books complicates assessments of measurement validity for cultural analytics by contaminating test data; we show that models perform much better on memorized books than on non-memorized books for downstream tasks. We argue that this supports a case for open models whose training data is known.", }

In this work, we carry out a data archaeology to infer books that are known to ChatGPT and GPT-4 using a name cloze membership inference query. We find that OpenAI models have memorized a wide collection of copyrighted materials, and that the degree of memorization is tied to the frequency with which passages of those books appear on the web. The ability of these models to memorize an unknown set of books complicates assessments of measurement validity for cultural analytics by contaminating test data; we show that models perform much better on memorized books than on non-memorized books for downstream tasks. We argue that this supports a case for open models whose training data is known.

Speak, Memory: An Archaeology of Books Known to ChatGPT/GPT-4

emnlp-main.453

2305.00118

Poster

https://aclanthology.org/2023.emnlp-main.454.bib

https://aclanthology.org/2023.emnlp-main.454/

@inproceedings{di-marco-etal-2023-study, title = "A Study on Accessing Linguistic Information in Pre-Trained Language Models by Using Prompts", author = {Di Marco, Marion and H{\"a}mmerl, Katharina and Fraser, Alexander}, editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.454", doi = "10.18653/v1/2023.emnlp-main.454", pages = "7328--7336", abstract = "We study whether linguistic information in pre-trained multilingual language models can be accessed by human language: So far, there is no easy method to directly obtain linguistic information and gain insights into the linguistic principles encoded in such models. We use the technique of prompting and formulate linguistic tasks to test the LM{'}s access to explicit grammatical principles and study how effective this method is at providing access to linguistic features. Our experiments on German, Icelandic and Spanish show that some linguistic properties can in fact be accessed through prompting, whereas others are harder to capture.", }

We study whether linguistic information in pre-trained multilingual language models can be accessed by human language: So far, there is no easy method to directly obtain linguistic information and gain insights into the linguistic principles encoded in such models. We use the technique of prompting and formulate linguistic tasks to test the LM{'}s access to explicit grammatical principles and study how effective this method is at providing access to linguistic features. Our experiments on German, Icelandic and Spanish show that some linguistic properties can in fact be accessed through prompting, whereas others are harder to capture.

A Study on Accessing Linguistic Information in Pre-Trained Language Models by Using Prompts

emnlp-main.454

Poster

https://aclanthology.org/2023.emnlp-main.455.bib

https://aclanthology.org/2023.emnlp-main.455/

@inproceedings{funkquist-etal-2023-citebench, title = "{C}ite{B}ench: A Benchmark for Scientific Citation Text Generation", author = "Funkquist, Martin and Kuznetsov, Ilia and Hou, Yufang and Gurevych, Iryna", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.455", doi = "10.18653/v1/2023.emnlp-main.455", pages = "7337--7353", abstract = "Science progresses by building upon the prior body of knowledge documented in scientific publications. The acceleration of research makes it hard to stay up-to-date with the recent developments and to summarize the ever-growing body of prior work. To address this, the task of citation text generation aims to produce accurate textual summaries given a set of papers-to-cite and the citing paper context. Due to otherwise rare explicit anchoring of cited documents in the citing paper, citation text generation provides an excellent opportunity to study how humans aggregate and synthesize textual knowledge from sources. Yet, existing studies are based upon widely diverging task definitions, which makes it hard to study this task systematically. To address this challenge, we propose CiteBench: a benchmark for citation text generation that unifies multiple diverse datasets and enables standardized evaluation of citation text generation models across task designs and domains. Using the new benchmark, we investigate the performance of multiple strong baselines, test their transferability between the datasets, and deliver new insights into the task definition and evaluation to guide future research in citation text generation. We make the code for CiteBench publicly available at https://github.com/UKPLab/citebench.", }

Science progresses by building upon the prior body of knowledge documented in scientific publications. The acceleration of research makes it hard to stay up-to-date with the recent developments and to summarize the ever-growing body of prior work. To address this, the task of citation text generation aims to produce accurate textual summaries given a set of papers-to-cite and the citing paper context. Due to otherwise rare explicit anchoring of cited documents in the citing paper, citation text generation provides an excellent opportunity to study how humans aggregate and synthesize textual knowledge from sources. Yet, existing studies are based upon widely diverging task definitions, which makes it hard to study this task systematically. To address this challenge, we propose CiteBench: a benchmark for citation text generation that unifies multiple diverse datasets and enables standardized evaluation of citation text generation models across task designs and domains. Using the new benchmark, we investigate the performance of multiple strong baselines, test their transferability between the datasets, and deliver new insights into the task definition and evaluation to guide future research in citation text generation. We make the code for CiteBench publicly available at https://github.com/UKPLab/citebench.

CiteBench: A Benchmark for Scientific Citation Text Generation

emnlp-main.455

2212.09577

Poster

https://aclanthology.org/2023.emnlp-main.456.bib

https://aclanthology.org/2023.emnlp-main.456/

@inproceedings{zhang-etal-2023-heuristic, title = "From Heuristic to Analytic: Cognitively Motivated Strategies for Coherent Physical Commonsense Reasoning", author = "Zhang, Zheyuan and Storks, Shane and Hu, Fengyuan and Sohn, Sungryull and Lee, Moontae and Lee, Honglak and Chai, Joyce", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.456", doi = "10.18653/v1/2023.emnlp-main.456", pages = "7354--7379", abstract = "Pre-trained language models (PLMs) have shown impressive performance in various language tasks. However, they are prone to spurious correlations, and often generate illusory information. In real-world applications, PLMs should justify decisions with formalized, coherent reasoning chains, but this challenge remains under-explored. Cognitive psychology theorizes that humans are capable of utilizing fast and intuitive *heuristic* thinking to make decisions based on past experience, then rationalizing the decisions through slower and deliberative *analytic* reasoning. We incorporate these interlinked dual processes in fine-tuning and in-context learning with PLMs, applying them to two language understanding tasks that require coherent physical commonsense reasoning. We show that our proposed Heuristic-Analytic Reasoning (HAR) strategies drastically improve the coherence of rationalizations for model decisions, yielding state-of-the-art results on Tiered Reasoning for Intuitive Physics (TRIP). We also find that this improved coherence is a direct result of more faithful attention to relevant language context in each step of reasoning. Our findings suggest that human-like reasoning strategies can effectively improve the coherence and reliability of PLM reasoning.", }

Pre-trained language models (PLMs) have shown impressive performance in various language tasks. However, they are prone to spurious correlations, and often generate illusory information. In real-world applications, PLMs should justify decisions with formalized, coherent reasoning chains, but this challenge remains under-explored. Cognitive psychology theorizes that humans are capable of utilizing fast and intuitive *heuristic* thinking to make decisions based on past experience, then rationalizing the decisions through slower and deliberative *analytic* reasoning. We incorporate these interlinked dual processes in fine-tuning and in-context learning with PLMs, applying them to two language understanding tasks that require coherent physical commonsense reasoning. We show that our proposed Heuristic-Analytic Reasoning (HAR) strategies drastically improve the coherence of rationalizations for model decisions, yielding state-of-the-art results on Tiered Reasoning for Intuitive Physics (TRIP). We also find that this improved coherence is a direct result of more faithful attention to relevant language context in each step of reasoning. Our findings suggest that human-like reasoning strategies can effectively improve the coherence and reliability of PLM reasoning.

From Heuristic to Analytic: Cognitively Motivated Strategies for Coherent Physical Commonsense Reasoning

emnlp-main.456

2310.18364

https://huggingface.co/papers/2310.18364

Poster

https://aclanthology.org/2023.emnlp-main.457.bib

https://aclanthology.org/2023.emnlp-main.457/

@inproceedings{kudo-etal-2023-challenging, title = "A Challenging Multimodal Video Summary: Simultaneously Extracting and Generating Keyframe-Caption Pairs from Video", author = "Kudo, Keito and Nagasawa, Haruki and Suzuki, Jun and Shimizu, Nobuyuki", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.457", doi = "10.18653/v1/2023.emnlp-main.457", pages = "7380--7402", abstract = "This paper proposes a practical multimodal video summarization task setting and a dataset to train and evaluate the task. The target task involves summarizing a given video into a predefined number of keyframe-caption pairs and displaying them in a listable format to grasp the video content quickly. This task aims to extract crucial scenes from the video in the form of images (keyframes) and generate corresponding captions explaining each keyframe{'}s situation. This task is useful as a practical application and presents a highly challenging problem worthy of study. Specifically, achieving simultaneous optimization of the keyframe selection performance and caption quality necessitates careful consideration of the mutual dependence on both preceding and subsequent keyframes and captions. To facilitate subsequent research in this field, we also construct a dataset by expanding upon existing datasets and propose an evaluation framework. Furthermore, we develop two baseline systems and report their respective performance.", }

This paper proposes a practical multimodal video summarization task setting and a dataset to train and evaluate the task. The target task involves summarizing a given video into a predefined number of keyframe-caption pairs and displaying them in a listable format to grasp the video content quickly. This task aims to extract crucial scenes from the video in the form of images (keyframes) and generate corresponding captions explaining each keyframe{'}s situation. This task is useful as a practical application and presents a highly challenging problem worthy of study. Specifically, achieving simultaneous optimization of the keyframe selection performance and caption quality necessitates careful consideration of the mutual dependence on both preceding and subsequent keyframes and captions. To facilitate subsequent research in this field, we also construct a dataset by expanding upon existing datasets and propose an evaluation framework. Furthermore, we develop two baseline systems and report their respective performance.

A Challenging Multimodal Video Summary: Simultaneously Extracting and Generating Keyframe-Caption Pairs from Video

emnlp-main.457

2312.01575

https://huggingface.co/papers/2312.01575

Poster

https://aclanthology.org/2023.emnlp-main.458.bib

https://aclanthology.org/2023.emnlp-main.458/

@inproceedings{karamolegkou-etal-2023-copyright, title = "Copyright Violations and Large Language Models", author = "Karamolegkou, Antonia and Li, Jiaang and Zhou, Li and S{\o}gaard, Anders", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.458", doi = "10.18653/v1/2023.emnlp-main.458", pages = "7403--7412", abstract = "Language models may memorize more than just facts, including entire chunks of texts seen during training. Fair use exemptions to copyright laws typically allow for limited use of copyrighted material without permission from the copyright holder, but typically for extraction of information from copyrighted materials, rather than \textit{verbatim} reproduction. This work explores the issue of copyright violations and large language models through the lens of verbatim memorization, focusing on possible redistribution of copyrighted text. We present experiments with a range of language models over a collection of popular books and coding problems, providing a conservative characterization of the extent to which language models can redistribute these materials. Overall, this research highlights the need for further examination and the potential impact on future developments in natural language processing to ensure adherence to copyright regulations. Code is at https://github.com/coastalcph/CopyrightLLMs.", }

Language models may memorize more than just facts, including entire chunks of texts seen during training. Fair use exemptions to copyright laws typically allow for limited use of copyrighted material without permission from the copyright holder, but typically for extraction of information from copyrighted materials, rather than \textit{verbatim} reproduction. This work explores the issue of copyright violations and large language models through the lens of verbatim memorization, focusing on possible redistribution of copyrighted text. We present experiments with a range of language models over a collection of popular books and coding problems, providing a conservative characterization of the extent to which language models can redistribute these materials. Overall, this research highlights the need for further examination and the potential impact on future developments in natural language processing to ensure adherence to copyright regulations. Code is at https://github.com/coastalcph/CopyrightLLMs.

Copyright Violations and Large Language Models

emnlp-main.458

2310.13771

https://huggingface.co/papers/2310.13771

Oral

https://aclanthology.org/2023.emnlp-main.459.bib

https://aclanthology.org/2023.emnlp-main.459/

@inproceedings{hou-etal-2023-effects, title = "Effects of sub-word segmentation on performance of transformer language models", author = "Hou, Jue and Katinskaia, Anisia and Vu, Anh-Duc and Yangarber, Roman", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.459", doi = "10.18653/v1/2023.emnlp-main.459", pages = "7413--7425", abstract = "Language modeling is a fundamental task in natural language processing, which has been thoroughly explored with various architectures and hyperparameters. However, few studies focus on the effect of sub-word segmentation on the performance of language models (LMs). In this paper, we compare GPT and BERT models trained with the statistical segmentation algorithm BPE vs. two unsupervised algorithms for morphological segmentation {---} Morfessor and StateMorph. We train the models for several languages {---} including ones with very rich morphology {---} and compare their performance with different segmentation algorithms, vocabulary sizes, and model sizes. The results show that training with morphological segmentation allows the LMs to: (1) achieve lower perplexity, (2) converge more efficiently in terms of training time, and (3) achieve equivalent or better evaluation scores on downstream tasks. Lastly, we show that (4) LMs of smaller size using morphological segmentation can perform comparably to models of larger size trained with BPE {---} both in terms of (1) perplexity and (3) scores on downstream tasks. Points (2) and (4) impact on sustainability, since they reduce the model cost; and while 2 reduces cost only in the training phase, 4 does so also in the inference phase.", }

Language modeling is a fundamental task in natural language processing, which has been thoroughly explored with various architectures and hyperparameters. However, few studies focus on the effect of sub-word segmentation on the performance of language models (LMs). In this paper, we compare GPT and BERT models trained with the statistical segmentation algorithm BPE vs. two unsupervised algorithms for morphological segmentation {---} Morfessor and StateMorph. We train the models for several languages {---} including ones with very rich morphology {---} and compare their performance with different segmentation algorithms, vocabulary sizes, and model sizes. The results show that training with morphological segmentation allows the LMs to: (1) achieve lower perplexity, (2) converge more efficiently in terms of training time, and (3) achieve equivalent or better evaluation scores on downstream tasks. Lastly, we show that (4) LMs of smaller size using morphological segmentation can perform comparably to models of larger size trained with BPE {---} both in terms of (1) perplexity and (3) scores on downstream tasks. Points (2) and (4) impact on sustainability, since they reduce the model cost; and while 2 reduces cost only in the training phase, 4 does so also in the inference phase.

Effects of sub-word segmentation on performance of transformer language models

emnlp-main.459

2305.05480

Poster

https://aclanthology.org/2023.emnlp-main.460.bib

https://aclanthology.org/2023.emnlp-main.460/

@inproceedings{chiu-etal-2023-symbolic-planning, title = "Symbolic Planning and Code Generation for Grounded Dialogue", author = "Chiu, Justin and Zhao, Wenting and Chen, Derek and Vaduguru, Saujas and Rush, Alexander and Fried, Daniel", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.460", doi = "10.18653/v1/2023.emnlp-main.460", pages = "7426--7436", abstract = "Large language models (LLMs) excel at processing and generating text and code. However, LLMs have had limited applicability in grounded task-oriented dialogue as they are difficult to steer toward task objectives and fail to handle novel grounding. We present a modular and interpretable grounded dialogue system that addresses these shortcomings by composing LLMs with a symbolic planner and grounded code execution. Our system, consists of a reader and planner: the reader leverages an LLM to convert partner utterances into executable code, calling functions that perform grounding. The translated code{'}s output is stored to track dialogue state, while a symbolic planner determines the next appropriate response. We evaluate our system{'}s performance on the demanding OneCommon dialogue task, involving collaborative reference resolution on abstract images of scattered dots. Our system substantially outperforms the previous state-of-the-art, including improving task success in human evaluations from 56{\%} to 69{\%} in the most challenging setting.", }

Large language models (LLMs) excel at processing and generating text and code. However, LLMs have had limited applicability in grounded task-oriented dialogue as they are difficult to steer toward task objectives and fail to handle novel grounding. We present a modular and interpretable grounded dialogue system that addresses these shortcomings by composing LLMs with a symbolic planner and grounded code execution. Our system, consists of a reader and planner: the reader leverages an LLM to convert partner utterances into executable code, calling functions that perform grounding. The translated code{'}s output is stored to track dialogue state, while a symbolic planner determines the next appropriate response. We evaluate our system{'}s performance on the demanding OneCommon dialogue task, involving collaborative reference resolution on abstract images of scattered dots. Our system substantially outperforms the previous state-of-the-art, including improving task success in human evaluations from 56{\%} to 69{\%} in the most challenging setting.

Symbolic Planning and Code Generation for Grounded Dialogue

emnlp-main.460

2310.17140

Poster

https://aclanthology.org/2023.emnlp-main.461.bib

https://aclanthology.org/2023.emnlp-main.461/

@inproceedings{wan-etal-2023-universal, title = "Universal Self-Adaptive Prompting", author = "Wan, Xingchen and Sun, Ruoxi and Nakhost, Hootan and Dai, Hanjun and Eisenschlos, Julian and Arik, Sercan and Pfister, Tomas", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.461", doi = "10.18653/v1/2023.emnlp-main.461", pages = "7437--7462", abstract = "A hallmark of modern large language models (LLMs) is their impressive general zero-shot and few-shot abilities, often elicited through in-context learning (ICL) via prompting. However, while highly coveted and being the most general, zero-shot performances in LLMs are still typically weaker due to the lack of guidance and the difficulty of applying existing automatic prompt design methods in general tasks when ground-truth labels are unavailable. In this study, we address this by presenting Universal Self-Adaptive Prompting (USP), an automatic prompt design approach specifically tailored for zero-shot learning (while compatible with few-shot). Requiring only a small amount of unlabeled data and an inference-only LLM, USP is highly versatile: to achieve universal prompting, USP categorizes a possible NLP task into one of the three possible task types and then uses a corresponding selector to select the most suitable queries and zero-shot model-generated responses as pseudo-demonstrations, thereby generalizing ICL to the zero-shot setup in a fully automated way. We evaluate USP with PaLM and PaLM 2 models and demonstrate performances that are considerably stronger than standard zero-shot baselines and often comparable to or even superior to few-shot baselines across more than 40 natural language understanding, natural language generation, and reasoning tasks.", }

A hallmark of modern large language models (LLMs) is their impressive general zero-shot and few-shot abilities, often elicited through in-context learning (ICL) via prompting. However, while highly coveted and being the most general, zero-shot performances in LLMs are still typically weaker due to the lack of guidance and the difficulty of applying existing automatic prompt design methods in general tasks when ground-truth labels are unavailable. In this study, we address this by presenting Universal Self-Adaptive Prompting (USP), an automatic prompt design approach specifically tailored for zero-shot learning (while compatible with few-shot). Requiring only a small amount of unlabeled data and an inference-only LLM, USP is highly versatile: to achieve universal prompting, USP categorizes a possible NLP task into one of the three possible task types and then uses a corresponding selector to select the most suitable queries and zero-shot model-generated responses as pseudo-demonstrations, thereby generalizing ICL to the zero-shot setup in a fully automated way. We evaluate USP with PaLM and PaLM 2 models and demonstrate performances that are considerably stronger than standard zero-shot baselines and often comparable to or even superior to few-shot baselines across more than 40 natural language understanding, natural language generation, and reasoning tasks.

Universal Self-Adaptive Prompting

emnlp-main.461

2305.14926

https://huggingface.co/papers/2305.14926

Poster

https://aclanthology.org/2023.emnlp-main.462.bib

https://aclanthology.org/2023.emnlp-main.462/

@inproceedings{badel-etal-2023-somali, title = "{S}omali Information Retrieval Corpus: Bridging the Gap between Query Translation and Dedicated Language Resources", author = "Badel, Abdisalam and Zhong, Ting and Tai, Wenxin and Zhou, Fan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.462", doi = "10.18653/v1/2023.emnlp-main.462", pages = "7463--7469", abstract = "Despite the growing use of the Somali language in various online domains, research on Somali language information retrieval remains limited and primarily relies on query translation due to the lack of a dedicated corpus. To address this problem, we collaborated with language experts and natural language processing (NLP) researchers to create an annotated corpus for Somali information retrieval. This corpus comprises 2335 documents collected from various well-known online sites, such as hiiraan online, dhacdo net, and Somali poetry books. We explain how the corpus was constructed, and develop a Somali language information retrieval system using a pseudo-relevance feedback (PRF) query expansion technique on the corpus. Note that collecting such a data set for the low-resourced Somali language can help overcome NLP barriers, such as the lack of electronically available data sets. Which, if available, can enable the development of various NLP tools and applications such as question-answering and text classification. It also provides researchers with a valuable resource for investigating and developing new techniques and approaches for Somali.", }

Despite the growing use of the Somali language in various online domains, research on Somali language information retrieval remains limited and primarily relies on query translation due to the lack of a dedicated corpus. To address this problem, we collaborated with language experts and natural language processing (NLP) researchers to create an annotated corpus for Somali information retrieval. This corpus comprises 2335 documents collected from various well-known online sites, such as hiiraan online, dhacdo net, and Somali poetry books. We explain how the corpus was constructed, and develop a Somali language information retrieval system using a pseudo-relevance feedback (PRF) query expansion technique on the corpus. Note that collecting such a data set for the low-resourced Somali language can help overcome NLP barriers, such as the lack of electronically available data sets. Which, if available, can enable the development of various NLP tools and applications such as question-answering and text classification. It also provides researchers with a valuable resource for investigating and developing new techniques and approaches for Somali.

Somali Information Retrieval Corpus: Bridging the Gap between Query Translation and Dedicated Language Resources

emnlp-main.462

Poster

https://aclanthology.org/2023.emnlp-main.463.bib

https://aclanthology.org/2023.emnlp-main.463/

@inproceedings{zhu-etal-2023-beat, title = "Beat {LLM}s at Their Own Game: Zero-Shot {LLM}-Generated Text Detection via Querying {C}hat{GPT}", author = "Zhu, Biru and Yuan, Lifan and Cui, Ganqu and Chen, Yangyi and Fu, Chong and He, Bingxiang and Deng, Yangdong and Liu, Zhiyuan and Sun, Maosong and Gu, Ming", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.463", doi = "10.18653/v1/2023.emnlp-main.463", pages = "7470--7483", abstract = "Large language models (LLMs), e.g., ChatGPT, have revolutionized the domain of natural language processing because of their excellent performance on various tasks. Despite their great potential, LLMs also incur serious concerns as they are likely to be misused. There are already reported cases of academic cheating by using LLMs. Thus, it is a pressing problem to identify LLM-generated texts. In this work, we design a zero-shot black-box method for detecting LLM-generated texts. The key idea is to revise the text to be detected using the ChatGPT model. Our method is based on the intuition that the ChatGPT model will make fewer revisions to LLM-generated texts than it does to human-written texts, because the texts generated by LLMs are more in accord with the generation logic and statistical patterns learned by LLMs like ChatGPT. Thus, if the text to be detected and its ChatGPT-revised version have a higher degree of similarity, the text is more likely to be LLM-generated. Extensive experiments on various datasets and tasks show that our method can effectively detect LLM-generated texts. Moreover, compared with other detection methods, our method has better generalization ability and is more stable across various datasets. The codes are publicly available at https://github.com/thunlp/LLM-generated-text-detection.", }

Large language models (LLMs), e.g., ChatGPT, have revolutionized the domain of natural language processing because of their excellent performance on various tasks. Despite their great potential, LLMs also incur serious concerns as they are likely to be misused. There are already reported cases of academic cheating by using LLMs. Thus, it is a pressing problem to identify LLM-generated texts. In this work, we design a zero-shot black-box method for detecting LLM-generated texts. The key idea is to revise the text to be detected using the ChatGPT model. Our method is based on the intuition that the ChatGPT model will make fewer revisions to LLM-generated texts than it does to human-written texts, because the texts generated by LLMs are more in accord with the generation logic and statistical patterns learned by LLMs like ChatGPT. Thus, if the text to be detected and its ChatGPT-revised version have a higher degree of similarity, the text is more likely to be LLM-generated. Extensive experiments on various datasets and tasks show that our method can effectively detect LLM-generated texts. Moreover, compared with other detection methods, our method has better generalization ability and is more stable across various datasets. The codes are publicly available at https://github.com/thunlp/LLM-generated-text-detection.

Beat LLMs at Their Own Game: Zero-Shot LLM-Generated Text Detection via Querying ChatGPT

emnlp-main.463

Poster

https://aclanthology.org/2023.emnlp-main.464.bib

https://aclanthology.org/2023.emnlp-main.464/

@inproceedings{hu-etal-2023-faithful, title = "Faithful Model Evaluation for Model-Based Metrics", author = "Hu, Qian and Goyal, Palash and Gupta, Rahul", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.464", doi = "10.18653/v1/2023.emnlp-main.464", pages = "7484--7489", abstract = "Statistical significance testing is used in natural language processing (NLP) to determine whether the results of a study or experiment are likely to be due to chance or if they reflect a genuine relationship. A key step in significance testing is the estimation of confidence interval which is a function of sample variance. Sample variance calculation is straightforward when evaluating against ground truth. However, in many cases, a metric model is often used for evaluation. For example, to compare toxicity of two large language models, a toxicity classifier is used for evaluation. Existing works usually do not consider the variance change due to metric model errors, which can lead to wrong conclusions. In this work, we establish the mathematical foundation of significance testing for model-based metrics. With experiments on public benchmark datasets and a production system, we show that considering metric model errors to calculate sample variances for model-based metrics changes the conclusions in certain experiments.", }

Statistical significance testing is used in natural language processing (NLP) to determine whether the results of a study or experiment are likely to be due to chance or if they reflect a genuine relationship. A key step in significance testing is the estimation of confidence interval which is a function of sample variance. Sample variance calculation is straightforward when evaluating against ground truth. However, in many cases, a metric model is often used for evaluation. For example, to compare toxicity of two large language models, a toxicity classifier is used for evaluation. Existing works usually do not consider the variance change due to metric model errors, which can lead to wrong conclusions. In this work, we establish the mathematical foundation of significance testing for model-based metrics. With experiments on public benchmark datasets and a production system, we show that considering metric model errors to calculate sample variances for model-based metrics changes the conclusions in certain experiments.

Faithful Model Evaluation for Model-Based Metrics

emnlp-main.464

2312.17254

Poster

https://aclanthology.org/2023.emnlp-main.465.bib

https://aclanthology.org/2023.emnlp-main.465/

@inproceedings{zhang-etal-2023-content, title = "Content- and Topology-Aware Representation Learning for Scientific Multi-Literature", author = "Zhang, Kai and Song, Kaisong and Kang, Yangyang and Liu, Xiaozhong", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.465", doi = "10.18653/v1/2023.emnlp-main.465", pages = "7490--7502", abstract = "Representation learning forms an essential building block in the development of natural language processing architectures. To date, mainstream approaches focus on learning textual information at the sentence- or document-level, unfortunately, overlooking the inter-document connections. This omission decreases the potency of downstream applications, particularly in multi-document settings. To address this issue, embeddings equipped with latent semantic and rich relatedness information are needed. In this paper, we propose SMRC$^{2}$, which extends representation learning to the multi-document level. Our model jointly learns latent semantic information from content and rich relatedness information from topological networks. Unlike previous studies, our work takes multi-document as input and integrates both semantic and relatedness information using a shared space via language model and graph structure. Our extensive experiments confirm the superiority and effectiveness of our approach. To encourage further research in scientific multi-literature representation learning, we will release our code and a new dataset from the biomedical domain.", }

Representation learning forms an essential building block in the development of natural language processing architectures. To date, mainstream approaches focus on learning textual information at the sentence- or document-level, unfortunately, overlooking the inter-document connections. This omission decreases the potency of downstream applications, particularly in multi-document settings. To address this issue, embeddings equipped with latent semantic and rich relatedness information are needed. In this paper, we propose SMRC$^{2}$, which extends representation learning to the multi-document level. Our model jointly learns latent semantic information from content and rich relatedness information from topological networks. Unlike previous studies, our work takes multi-document as input and integrates both semantic and relatedness information using a shared space via language model and graph structure. Our extensive experiments confirm the superiority and effectiveness of our approach. To encourage further research in scientific multi-literature representation learning, we will release our code and a new dataset from the biomedical domain.

Content- and Topology-Aware Representation Learning for Scientific Multi-Literature

emnlp-main.465

Poster

https://aclanthology.org/2023.emnlp-main.466.bib

https://aclanthology.org/2023.emnlp-main.466/

@inproceedings{wilcox-etal-2023-language, title = "Language Model Quality Correlates with Psychometric Predictive Power in Multiple Languages", author = "Wilcox, Ethan and Meister, Clara and Cotterell, Ryan and Pimentel, Tiago", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.466", doi = "10.18653/v1/2023.emnlp-main.466", pages = "7503--7511", abstract = "Surprisal theory (Hale, 2001; Levy, 2008) posits that a word{'}s reading time is proportional to its surprisal (i.e., to its negative log probability given the proceeding context). Since we are unable to access a word{'}s ground-truth probability, surprisal theory has been empirically tested using surprisal estimates from language models (LMs). Under the premise that surprisal theory holds, we would expect that higher quality language models provide more powerful predictors of human reading behavior{---}a conjecture we dub the quality{--}power (QP) hypothesis. Unfortunately, empirical support for the QP hypothesis is mixed. Some studies in English have found correlations between LM quality and predictive power, but other studies using Japanese data, as well as using larger English LMs, find no such correlations. In this work, we conduct a systematic crosslinguistic assessment of the QP hypothesis. We train LMs from scratch on small- and medium-sized datasets from 13 languages (across five language families) and assess their ability to predict eye tracking data. We find correlations between LM quality and power in eleven of these thirteen languages, suggesting that, within the range of model classes and sizes tested, better language models are indeed better predictors of human language processing behaviors.", }

Surprisal theory (Hale, 2001; Levy, 2008) posits that a word{'}s reading time is proportional to its surprisal (i.e., to its negative log probability given the proceeding context). Since we are unable to access a word{'}s ground-truth probability, surprisal theory has been empirically tested using surprisal estimates from language models (LMs). Under the premise that surprisal theory holds, we would expect that higher quality language models provide more powerful predictors of human reading behavior{---}a conjecture we dub the quality{--}power (QP) hypothesis. Unfortunately, empirical support for the QP hypothesis is mixed. Some studies in English have found correlations between LM quality and predictive power, but other studies using Japanese data, as well as using larger English LMs, find no such correlations. In this work, we conduct a systematic crosslinguistic assessment of the QP hypothesis. We train LMs from scratch on small- and medium-sized datasets from 13 languages (across five language families) and assess their ability to predict eye tracking data. We find correlations between LM quality and power in eleven of these thirteen languages, suggesting that, within the range of model classes and sizes tested, better language models are indeed better predictors of human language processing behaviors.

Language Model Quality Correlates with Psychometric Predictive Power in Multiple Languages

emnlp-main.466

Poster

https://aclanthology.org/2023.emnlp-main.467.bib

https://aclanthology.org/2023.emnlp-main.467/

@inproceedings{yan-etal-2023-joint, title = "Joint Entity and Relation Extraction with Span Pruning and Hypergraph Neural Networks", author = "Yan, Zhaohui and Yang, Songlin and Liu, Wei and Tu, Kewei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.467", doi = "10.18653/v1/2023.emnlp-main.467", pages = "7512--7526", abstract = "Entity and Relation Extraction (ERE) is an important task in information extraction. Recent marker-based pipeline models achieve state-of-the-art performance, but still suffer from the error propagation issue. Also, most of current ERE models do not take into account higher-order interactions between multiple entities and relations, while higher-order modeling could be beneficial.In this work, we propose HyperGraph neural network for ERE (HGERE), which is built upon the PL-marker (a state-of-the-art marker-based pipleline model). To alleviate error propagation, we use a high-recall pruner mechanism to transfer the burden of entity identification and labeling from the NER module to the joint module of our model. For higher-order modeling, we build a hypergraph, where nodes are entities (provided by the span pruner) and relations thereof, and hyperedges encode interactions between two different relations or between a relation and its associated subject and object entities. We then run a hypergraph neural network for higher-order inference by applying message passing over the built hypergraph. Experiments on three widely used benchmarks (ACE2004, ACE2005 and SciERC) for ERE task show significant improvements over the previous state-of-the-art PL-marker.", }

Entity and Relation Extraction (ERE) is an important task in information extraction. Recent marker-based pipeline models achieve state-of-the-art performance, but still suffer from the error propagation issue. Also, most of current ERE models do not take into account higher-order interactions between multiple entities and relations, while higher-order modeling could be beneficial.In this work, we propose HyperGraph neural network for ERE (HGERE), which is built upon the PL-marker (a state-of-the-art marker-based pipleline model). To alleviate error propagation, we use a high-recall pruner mechanism to transfer the burden of entity identification and labeling from the NER module to the joint module of our model. For higher-order modeling, we build a hypergraph, where nodes are entities (provided by the span pruner) and relations thereof, and hyperedges encode interactions between two different relations or between a relation and its associated subject and object entities. We then run a hypergraph neural network for higher-order inference by applying message passing over the built hypergraph. Experiments on three widely used benchmarks (ACE2004, ACE2005 and SciERC) for ERE task show significant improvements over the previous state-of-the-art PL-marker.

Joint Entity and Relation Extraction with Span Pruning and Hypergraph Neural Networks

emnlp-main.467

2310.17238

Poster

https://aclanthology.org/2023.emnlp-main.468.bib

https://aclanthology.org/2023.emnlp-main.468/

@inproceedings{arora-etal-2023-llms, title = "Have {LLM}s Advanced Enough? A Challenging Problem Solving Benchmark For Large Language Models", author = "Arora, Daman and Singh, Himanshu and {Mausam}", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.468", doi = "10.18653/v1/2023.emnlp-main.468", pages = "7527--7543", abstract = "The performance of large language models (LLMs) on existing reasoning benchmarks has significantly improved over the past years. In response, we present JEEBench, a considerably more challenging benchmark dataset for evaluating the problem solving abilities of LLMs. We curate 515 challenging pre-engineering mathematics, physics and chemistry problems from the highly competitive IIT JEE-Advanced exam. Long-horizon reasoning on top of deep in-domain knowledge is essential for solving problems in this benchmark. Our evaluation on various open-source and proprietary models reveals that the highest performance, even after using techniques like self-consistency, self-refinement and chain-of-thought prompting, is less than 40{\%}. The typical failure modes of GPT-4, the best model, are errors in algebraic manipulation, difficulty in grounding abstract concepts into mathematical equations accurately and failure in retrieving relevant domain-specific concepts. We also observe that by mere prompting, GPT-4 is unable to assess risk introduced by negative marking for incorrect answers. For this, we develop a post-hoc confidence-thresholding method over self-consistency, which enables effective response selection. We hope that our challenging benchmark will guide future re-search in problem-solving using LLMs.", }

The performance of large language models (LLMs) on existing reasoning benchmarks has significantly improved over the past years. In response, we present JEEBench, a considerably more challenging benchmark dataset for evaluating the problem solving abilities of LLMs. We curate 515 challenging pre-engineering mathematics, physics and chemistry problems from the highly competitive IIT JEE-Advanced exam. Long-horizon reasoning on top of deep in-domain knowledge is essential for solving problems in this benchmark. Our evaluation on various open-source and proprietary models reveals that the highest performance, even after using techniques like self-consistency, self-refinement and chain-of-thought prompting, is less than 40{\%}. The typical failure modes of GPT-4, the best model, are errors in algebraic manipulation, difficulty in grounding abstract concepts into mathematical equations accurately and failure in retrieving relevant domain-specific concepts. We also observe that by mere prompting, GPT-4 is unable to assess risk introduced by negative marking for incorrect answers. For this, we develop a post-hoc confidence-thresholding method over self-consistency, which enables effective response selection. We hope that our challenging benchmark will guide future re-search in problem-solving using LLMs.

Have LLMs Advanced Enough? A Challenging Problem Solving Benchmark For Large Language Models

emnlp-main.468

Poster

https://aclanthology.org/2023.emnlp-main.469.bib

https://aclanthology.org/2023.emnlp-main.469/

@inproceedings{opper-etal-2023-strae, title = "{S}tr{AE}: Autoencoding for Pre-Trained Embeddings using Explicit Structure", author = "Opper, Mattia and Prokhorov, Victor and N, Siddharth", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.469", doi = "10.18653/v1/2023.emnlp-main.469", pages = "7544--7560", abstract = "This work presents StrAE: a Structured Autoencoder framework that through strict adherence to explicit structure, and use of a novel contrastive objective over tree-structured representations, enables effective learning of multi-level representations. Through comparison over different forms of structure, we verify that our results are directly attributable to the informativeness of the structure provided as input, and show that this is not the case for existing tree models. We then further extend StrAE to allow the model to define its own compositions using a simple localised-merge algorithm. This variant, called Self-StrAE, outperforms baselines that don{'}t involve explicit hierarchical compositions, and is comparable to models given informative structure (e.g. constituency parses). Our experiments are conducted in a data-constrained (circa 10M tokens) setting to help tease apart the contribution of the inductive bias to effective learning. However, we find that this framework can be robust to scale, and when extended to a much larger dataset (circa 100M tokens), our 430 parameter model performs comparably to a 6-layer RoBERTa many orders of magnitude larger in size. Our findings support the utility of incorporating explicit composition as an inductive bias for effective representation learning.", }

This work presents StrAE: a Structured Autoencoder framework that through strict adherence to explicit structure, and use of a novel contrastive objective over tree-structured representations, enables effective learning of multi-level representations. Through comparison over different forms of structure, we verify that our results are directly attributable to the informativeness of the structure provided as input, and show that this is not the case for existing tree models. We then further extend StrAE to allow the model to define its own compositions using a simple localised-merge algorithm. This variant, called Self-StrAE, outperforms baselines that don{'}t involve explicit hierarchical compositions, and is comparable to models given informative structure (e.g. constituency parses). Our experiments are conducted in a data-constrained (circa 10M tokens) setting to help tease apart the contribution of the inductive bias to effective learning. However, we find that this framework can be robust to scale, and when extended to a much larger dataset (circa 100M tokens), our 430 parameter model performs comparably to a 6-layer RoBERTa many orders of magnitude larger in size. Our findings support the utility of incorporating explicit composition as an inductive bias for effective representation learning.

StrAE: Autoencoding for Pre-Trained Embeddings using Explicit Structure

emnlp-main.469

2305.05588

Poster

https://aclanthology.org/2023.emnlp-main.470.bib

https://aclanthology.org/2023.emnlp-main.470/

@inproceedings{kamoi-etal-2023-wice, title = "{W}i{CE}: Real-World Entailment for Claims in {W}ikipedia", author = "Kamoi, Ryo and Goyal, Tanya and Diego Rodriguez, Juan and Durrett, Greg", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.470", doi = "10.18653/v1/2023.emnlp-main.470", pages = "7561--7583", abstract = "Textual entailment models are increasingly applied in settings like fact-checking, presupposition verification in question answering, or summary evaluation. However, these represent a significant domain shift from existing entailment datasets, and models underperform as a result. We propose WiCE, a new fine-grained textual entailment dataset built on natural claim and evidence pairs extracted from Wikipedia. In addition to standard claim-level entailment, WiCE provides entailment judgments over sub-sentence units of the claim, and a minimal subset of evidence sentences that support each subclaim. To support this, we propose an automatic claim decomposition strategy using GPT-3.5 which we show is also effective at improving entailment models{'} performance on multiple datasets at test time. Finally, we show that real claims in our dataset involve challenging verification and retrieval problems that existing models fail to address.", }

Textual entailment models are increasingly applied in settings like fact-checking, presupposition verification in question answering, or summary evaluation. However, these represent a significant domain shift from existing entailment datasets, and models underperform as a result. We propose WiCE, a new fine-grained textual entailment dataset built on natural claim and evidence pairs extracted from Wikipedia. In addition to standard claim-level entailment, WiCE provides entailment judgments over sub-sentence units of the claim, and a minimal subset of evidence sentences that support each subclaim. To support this, we propose an automatic claim decomposition strategy using GPT-3.5 which we show is also effective at improving entailment models{'} performance on multiple datasets at test time. Finally, we show that real claims in our dataset involve challenging verification and retrieval problems that existing models fail to address.

WiCE: Real-World Entailment for Claims in Wikipedia

emnlp-main.470

2303.01432

https://huggingface.co/papers/2303.01432

Oral

https://aclanthology.org/2023.emnlp-main.471.bib

https://aclanthology.org/2023.emnlp-main.471/

@inproceedings{basit-etal-2023-natural, title = "Natural Disaster Tweets Classification Using Multimodal Data", author = "Basit, Mohammad and Alam, Bashir and Fatima, Zubaida and Shaikh, Salman", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.471", doi = "10.18653/v1/2023.emnlp-main.471", pages = "7584--7594", abstract = "Social media platforms are extensively used for expressing opinions or conveying information. The information available on such platforms can be used for various humanitarian and disaster-related tasks as distributing messages in different formats through social media is quick and easy. Often this useful information during disaster events goes to waste as efficient systems don{'}t exist which can turn these unstructured data into meaningful format which can ultimately assist aid agencies. In disaster identification and assessment, information available is naturally multimodal, however, most existing work has been solely focused on single modalities e.g. images or texts separately. When information from different modalities are integrated , it produces significantly better results. In this paper, we have explored different models which can lead to the development of a system that deals with multimodal datasets and can perform sequential hierarchical classification. Specifically, we aim to find the damage and its severity along with classifying the data into humanitarian categories. The different stages in the hierarchical classification have had their respective models selected by researching with many different modality specific models and approaches of multimodal classification including multi task learning. The hierarchical model can give results at different abstraction levels according to the use cases. Through extensive quantitative and qualitative analysis, we show how our system is effective in classifying the multimodal tweets along with an excellent computational efficiency and assessment performance. With the help of our approach, we aim to support disaster management through identification of situations involving humanitarian tragedies and aid in assessing the severity and type of damage.", }

Social media platforms are extensively used for expressing opinions or conveying information. The information available on such platforms can be used for various humanitarian and disaster-related tasks as distributing messages in different formats through social media is quick and easy. Often this useful information during disaster events goes to waste as efficient systems don{'}t exist which can turn these unstructured data into meaningful format which can ultimately assist aid agencies. In disaster identification and assessment, information available is naturally multimodal, however, most existing work has been solely focused on single modalities e.g. images or texts separately. When information from different modalities are integrated , it produces significantly better results. In this paper, we have explored different models which can lead to the development of a system that deals with multimodal datasets and can perform sequential hierarchical classification. Specifically, we aim to find the damage and its severity along with classifying the data into humanitarian categories. The different stages in the hierarchical classification have had their respective models selected by researching with many different modality specific models and approaches of multimodal classification including multi task learning. The hierarchical model can give results at different abstraction levels according to the use cases. Through extensive quantitative and qualitative analysis, we show how our system is effective in classifying the multimodal tweets along with an excellent computational efficiency and assessment performance. With the help of our approach, we aim to support disaster management through identification of situations involving humanitarian tragedies and aid in assessing the severity and type of damage.

Natural Disaster Tweets Classification Using Multimodal Data

emnlp-main.471

Poster

https://aclanthology.org/2023.emnlp-main.472.bib

https://aclanthology.org/2023.emnlp-main.472/

@inproceedings{pozzobon-etal-2023-challenges, title = "On the Challenges of Using Black-Box {API}s for Toxicity Evaluation in Research", author = "Pozzobon, Luiza and Ermis, Beyza and Lewis, Patrick and Hooker, Sara", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.472", doi = "10.18653/v1/2023.emnlp-main.472", pages = "7595--7609", abstract = "Perception of toxicity evolves over time and often differs between geographies and cultural backgrounds. Similarly, black-box commercially available APIs for detecting toxicity, such as the Perspective API, are not static, but frequently retrained to address any unattended weaknesses and biases. We evaluate the implications of these changes on the reproducibility of findings that compare the relative merits of models and methods that aim to curb toxicity. Our findings suggest that research that relied on inherited automatic toxicity scores to compare models and techniques may have resulted in inaccurate findings. Rescoring all models from HELM, a widely respected living benchmark, for toxicity with the recent version of the API led to a different ranking of widely used foundation models. We suggest caution in applying apples-to-apples comparisons between studies and call for a more structured approach to evaluating toxicity over time.", }

Perception of toxicity evolves over time and often differs between geographies and cultural backgrounds. Similarly, black-box commercially available APIs for detecting toxicity, such as the Perspective API, are not static, but frequently retrained to address any unattended weaknesses and biases. We evaluate the implications of these changes on the reproducibility of findings that compare the relative merits of models and methods that aim to curb toxicity. Our findings suggest that research that relied on inherited automatic toxicity scores to compare models and techniques may have resulted in inaccurate findings. Rescoring all models from HELM, a widely respected living benchmark, for toxicity with the recent version of the API led to a different ranking of widely used foundation models. We suggest caution in applying apples-to-apples comparisons between studies and call for a more structured approach to evaluating toxicity over time.

On the Challenges of Using Black-Box APIs for Toxicity Evaluation in Research

emnlp-main.472

2304.12397

https://huggingface.co/papers/2304.12397

Poster

https://aclanthology.org/2023.emnlp-main.473.bib

https://aclanthology.org/2023.emnlp-main.473/

@inproceedings{dinu-etal-2023-robocop, title = "{R}o{B}o{C}o{P}: A Comprehensive {RO}mance {BO}rrowing {CO}gnate Package and Benchmark for Multilingual Cognate Identification", author = "Dinu, Liviu and Uban, Ana and Cristea, Alina and Dinu, Anca and Iordache, Ioan-Bogdan and Georgescu, Simona and Zoicas, Laurentiu", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.473", doi = "10.18653/v1/2023.emnlp-main.473", pages = "7610--7629", abstract = "The identification of cognates is a fundamental process in historical linguistics, on which any further research is based. Even though there are several cognate databases for Romance languages, they are rather scattered, incomplete, noisy, contain unreliable information, or have uncertain availability. In this paper we introduce a comprehensive database of Romance cognates and borrowings based on the etymological information provided by the dictionaries. We extract pairs of cognates between any two Romance languages by parsing electronic dictionaries of Romanian, Italian, Spanish, Portuguese and French. Based on this resource, we propose a strong benchmark for the automatic detection of cognates, by applying machine learning and deep learning based methods on any two pairs of Romance languages. We find that automatic identification of cognates is possible with accuracy averaging around 94{\%} for the more difficult task formulations.", }

The identification of cognates is a fundamental process in historical linguistics, on which any further research is based. Even though there are several cognate databases for Romance languages, they are rather scattered, incomplete, noisy, contain unreliable information, or have uncertain availability. In this paper we introduce a comprehensive database of Romance cognates and borrowings based on the etymological information provided by the dictionaries. We extract pairs of cognates between any two Romance languages by parsing electronic dictionaries of Romanian, Italian, Spanish, Portuguese and French. Based on this resource, we propose a strong benchmark for the automatic detection of cognates, by applying machine learning and deep learning based methods on any two pairs of Romance languages. We find that automatic identification of cognates is possible with accuracy averaging around 94{\%} for the more difficult task formulations.

RoBoCoP: A Comprehensive ROmance BOrrowing COgnate Package and Benchmark for Multilingual Cognate Identification

emnlp-main.473

Poster

https://aclanthology.org/2023.emnlp-main.474.bib

https://aclanthology.org/2023.emnlp-main.474/

@inproceedings{wang-etal-2023-instructive, title = "Instructive Dialogue Summarization with Query Aggregations", author = "Wang, Bin and Liu, Zhengyuan and Chen, Nancy", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.474", doi = "10.18653/v1/2023.emnlp-main.474", pages = "7630--7653", abstract = "Conventional dialogue summarization methods directly generate summaries and do not consider user{'}s specific interests. This poses challenges in cases where the users are more focused on particular topics or aspects. With the advancement of instruction-finetuned language models, we introduce instruction-tuning to dialogues to expand the capability set of dialogue summarization models. To overcome the scarcity of instructive dialogue summarization data, we propose a three-step approach to synthesize high-quality query-based summarization triples. This process involves summary-anchored query generation, query filtering and query-based summary generation. By training a unified model called InstructDS (Instructive Dialogue Summarization) on three summarization datasets with multi-purpose instructive triples, we expand the capability of dialogue summarization models. We evaluate our method on four datasets, including dialogue summarization and dialogue reading comprehension. Experimental results show that our approach outperforms the state-of-the-art models and even models with larger sizes. Additionally, our model exhibits higher generalizability and faithfulness, as confirmed by human subjective evaluations.", }

Conventional dialogue summarization methods directly generate summaries and do not consider user{'}s specific interests. This poses challenges in cases where the users are more focused on particular topics or aspects. With the advancement of instruction-finetuned language models, we introduce instruction-tuning to dialogues to expand the capability set of dialogue summarization models. To overcome the scarcity of instructive dialogue summarization data, we propose a three-step approach to synthesize high-quality query-based summarization triples. This process involves summary-anchored query generation, query filtering and query-based summary generation. By training a unified model called InstructDS (Instructive Dialogue Summarization) on three summarization datasets with multi-purpose instructive triples, we expand the capability of dialogue summarization models. We evaluate our method on four datasets, including dialogue summarization and dialogue reading comprehension. Experimental results show that our approach outperforms the state-of-the-art models and even models with larger sizes. Additionally, our model exhibits higher generalizability and faithfulness, as confirmed by human subjective evaluations.

Instructive Dialogue Summarization with Query Aggregations

emnlp-main.474

2310.10981

https://huggingface.co/papers/2310.10981

Oral

https://aclanthology.org/2023.emnlp-main.475.bib

https://aclanthology.org/2023.emnlp-main.475/

@inproceedings{de-silva-etal-2023-semantic, title = "Semantic matching for text classification with complex class descriptions", author = "De Silva, Brian and Huang, Kuan-Wen and Lee, Gwang and Hovsepian, Karen and Xu, Yan and Shen, Mingwei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.475", doi = "10.18653/v1/2023.emnlp-main.475", pages = "7654--7680", abstract = "Text classifiers are an indispensable tool for machine learning practitioners, but adapting them to new classes is expensive. To reduce the cost of new classes, previous work exploits class descriptions and/or labels from existing classes. However, these approaches leave a gap in the model development cycle as they support either zero- or few-shot learning, but not both. Existing classifiers either do not work on zero-shot problems, or fail to improve much with few-shot labels. Further, prior work is aimed at concise class descriptions, which may be insufficient for complex classes. We overcome these shortcomings by casting text classification as a matching problem, where a model matches examples with relevant class descriptions. This formulation lets us leverage labels and complex class descriptions to perform zero- and few-shot learning on new classes. We compare this approach with numerous baselines on text classification tasks with complex class descriptions and find that it achieves strong zero-shot performance and scales well with few-shot samples, beating strong baselines by 22.48{\%} (average precision) in the 10-shot setting. Furthermore, we extend the popular Model-Agnostic Meta-Learning algorithm to the zero-shot matching setting and show it improves zero-shot performance by 4.29{\%}. Our results show that expressing text classification as a matching problem is a cost-effective way to address new classes. This strategy enables zero-shot learning for cold-start scenarios and few-shot learning so the model can improve until it is capable enough to deploy.", }

Text classifiers are an indispensable tool for machine learning practitioners, but adapting them to new classes is expensive. To reduce the cost of new classes, previous work exploits class descriptions and/or labels from existing classes. However, these approaches leave a gap in the model development cycle as they support either zero- or few-shot learning, but not both. Existing classifiers either do not work on zero-shot problems, or fail to improve much with few-shot labels. Further, prior work is aimed at concise class descriptions, which may be insufficient for complex classes. We overcome these shortcomings by casting text classification as a matching problem, where a model matches examples with relevant class descriptions. This formulation lets us leverage labels and complex class descriptions to perform zero- and few-shot learning on new classes. We compare this approach with numerous baselines on text classification tasks with complex class descriptions and find that it achieves strong zero-shot performance and scales well with few-shot samples, beating strong baselines by 22.48{\%} (average precision) in the 10-shot setting. Furthermore, we extend the popular Model-Agnostic Meta-Learning algorithm to the zero-shot matching setting and show it improves zero-shot performance by 4.29{\%}. Our results show that expressing text classification as a matching problem is a cost-effective way to address new classes. This strategy enables zero-shot learning for cold-start scenarios and few-shot learning so the model can improve until it is capable enough to deploy.

Semantic matching for text classification with complex class descriptions

emnlp-main.475

Poster

https://aclanthology.org/2023.emnlp-main.476.bib

https://aclanthology.org/2023.emnlp-main.476/

@inproceedings{gu-etal-2023-madnet, title = "{MADN}et: Maximizing Addressee Deduction Expectation for Multi-Party Conversation Generation", author = "Gu, Jia-Chen and Tan, Chao-Hong and Chu, Caiyuan and Ling, Zhen-Hua and Tao, Chongyang and Liu, Quan and Liu, Cong", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.476", doi = "10.18653/v1/2023.emnlp-main.476", pages = "7681--7692", abstract = "Modeling multi-party conversations (MPCs) with graph neural networks has been proven effective at capturing complicated and graphical information flows. However, existing methods rely heavily on the necessary addressee labels and can only be applied to an ideal setting where each utterance must be tagged with an {``}@{''} or other equivalent addressee label. To study the scarcity of addressee labels which is a common issue in MPCs, we propose MADNet that maximizes addressee deduction expectation in heterogeneous graph neural networks for MPC generation. Given an MPC with a few addressee labels missing, existing methods fail to build a consecutively connected conversation graph, but only a few separate conversation fragments instead. To ensure message passing between these conversation fragments, four additional types of latent edges are designed to complete a fully-connected graph. Besides, to optimize the edge-type-dependent message passing for those utterances without addressee labels, an Expectation-Maximization-based method that iteratively generates silver addressee labels (E step), and optimizes the quality of generated responses (M step), is designed. Experimental results on two Ubuntu IRC channel benchmarks show that MADNet outperforms various baseline models on the task of MPC generation, especially under the more common and challenging setting where part of addressee labels are missing.", }

Modeling multi-party conversations (MPCs) with graph neural networks has been proven effective at capturing complicated and graphical information flows. However, existing methods rely heavily on the necessary addressee labels and can only be applied to an ideal setting where each utterance must be tagged with an {``}@{''} or other equivalent addressee label. To study the scarcity of addressee labels which is a common issue in MPCs, we propose MADNet that maximizes addressee deduction expectation in heterogeneous graph neural networks for MPC generation. Given an MPC with a few addressee labels missing, existing methods fail to build a consecutively connected conversation graph, but only a few separate conversation fragments instead. To ensure message passing between these conversation fragments, four additional types of latent edges are designed to complete a fully-connected graph. Besides, to optimize the edge-type-dependent message passing for those utterances without addressee labels, an Expectation-Maximization-based method that iteratively generates silver addressee labels (E step), and optimizes the quality of generated responses (M step), is designed. Experimental results on two Ubuntu IRC channel benchmarks show that MADNet outperforms various baseline models on the task of MPC generation, especially under the more common and challenging setting where part of addressee labels are missing.

MADNet: Maximizing Addressee Deduction Expectation for Multi-Party Conversation Generation

emnlp-main.476

2305.12733

Oral

https://aclanthology.org/2023.emnlp-main.477.bib

https://aclanthology.org/2023.emnlp-main.477/

@inproceedings{lee-etal-2023-glen, title = "{GLEN}: Generative Retrieval via Lexical Index Learning", author = "Lee, Sunkyung and Choi, Minjin and Lee, Jongwuk", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.477", doi = "10.18653/v1/2023.emnlp-main.477", pages = "7693--7704", abstract = "Generative retrieval shed light on a new paradigm of document retrieval, aiming to directly generate the identifier of a relevant document for a query. While it takes advantage of bypassing the construction of auxiliary index structures, existing studies face two significant challenges: (i) the discrepancy between the knowledge of pre-trained language models and identifiers and (ii) the gap between training and inference that poses difficulty in learning to rank. To overcome these challenges, we propose a novel generative retrieval method, namely Generative retrieval via LExical iNdex learning (GLEN). For training, GLEN effectively exploits a dynamic lexical identifier using a two-phase index learning strategy, enabling it to learn meaningful lexical identifiers and relevance signals between queries and documents. For inference, GLEN utilizes collision-free inference, using identifier weights to rank documents without additional overhead. Experimental results prove that GLEN achieves state-of-the-art or competitive performance against existing generative retrieval methods on various benchmark datasets, e.g., NQ320k, MS MARCO, and BEIR. The code is available at https://github.com/skleee/GLEN.", }

Generative retrieval shed light on a new paradigm of document retrieval, aiming to directly generate the identifier of a relevant document for a query. While it takes advantage of bypassing the construction of auxiliary index structures, existing studies face two significant challenges: (i) the discrepancy between the knowledge of pre-trained language models and identifiers and (ii) the gap between training and inference that poses difficulty in learning to rank. To overcome these challenges, we propose a novel generative retrieval method, namely Generative retrieval via LExical iNdex learning (GLEN). For training, GLEN effectively exploits a dynamic lexical identifier using a two-phase index learning strategy, enabling it to learn meaningful lexical identifiers and relevance signals between queries and documents. For inference, GLEN utilizes collision-free inference, using identifier weights to rank documents without additional overhead. Experimental results prove that GLEN achieves state-of-the-art or competitive performance against existing generative retrieval methods on various benchmark datasets, e.g., NQ320k, MS MARCO, and BEIR. The code is available at https://github.com/skleee/GLEN.

GLEN: Generative Retrieval via Lexical Index Learning

emnlp-main.477

2311.03057

Poster

https://aclanthology.org/2023.emnlp-main.478.bib

https://aclanthology.org/2023.emnlp-main.478/

@inproceedings{zhang-etal-2023-turn, title = "Turn-Level Active Learning for Dialogue State Tracking", author = "Zhang, Zihan and Fang, Meng and Ye, Fanghua and Chen, Ling and Namazi-Rad, Mohammad-Reza", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.478", doi = "10.18653/v1/2023.emnlp-main.478", pages = "7705--7719", abstract = "Dialogue state tracking (DST) plays an important role in task-oriented dialogue systems. However, collecting a large amount of turn-by-turn annotated dialogue data is costly and inefficient. In this paper, we propose a novel turn-level active learning framework for DST to actively select turns in dialogues to annotate. Given the limited labelling budget, experimental results demonstrate the effectiveness of selective annotation of dialogue turns. Additionally, our approach can effectively achieve comparable DST performance to traditional training approaches with significantly less annotated data, which provides a more efficient way to annotate new dialogue data.", }

Dialogue state tracking (DST) plays an important role in task-oriented dialogue systems. However, collecting a large amount of turn-by-turn annotated dialogue data is costly and inefficient. In this paper, we propose a novel turn-level active learning framework for DST to actively select turns in dialogues to annotate. Given the limited labelling budget, experimental results demonstrate the effectiveness of selective annotation of dialogue turns. Additionally, our approach can effectively achieve comparable DST performance to traditional training approaches with significantly less annotated data, which provides a more efficient way to annotate new dialogue data.

Turn-Level Active Learning for Dialogue State Tracking

emnlp-main.478

2310.14513

Poster

https://aclanthology.org/2023.emnlp-main.479.bib

https://aclanthology.org/2023.emnlp-main.479/

@inproceedings{tu-etal-2023-resee, title = "{R}e{S}ee: Responding through Seeing Fine-grained Visual Knowledge in Open-domain Dialogue", author = "Tu, Haoqin and Li, Yitong and Mi, Fei and Yang, Zhongliang", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.479", doi = "10.18653/v1/2023.emnlp-main.479", pages = "7720--7735", abstract = "Incorporating visual knowledge into text-only dialogue systems has become a potential direction to imitate the way humans think, imagine, and communicate. However, existing multimodal dialogue systems are either confined by the scale and quality of available datasets or the coarse concept of visual knowledge. To address these issues, we provide a new paradigm of constructing multimodal dialogues as well as two datasets extended from text-only dialogues under such paradigm (ReSee-$\texttt{WoW}$, ReSee-$\texttt{DD}$). We propose to explicitly split the visual knowledge into finer granularity ({``}turn-level{''} and {``}entity-level{''}). To further boost the accuracy and diversity of augmented visual information, we retrieve them from the Internet or a large image dataset. To demonstrate the superiority and universality of the provided visual knowledge, we propose a simple but effective framework ReSee to add visual representation into vanilla dialogue models by modality concatenations. We also conduct extensive experiments and ablations w.r.t. different model configurations and visual knowledge settings. Empirical, encouraging results not only demonstrate the effectiveness of introducing visual knowledge at both entity and turn level but also verify the proposed model ReSee outperforms several state-of-the-art methods on automatic and human evaluations. By leveraging text and vision knowledge, ReSee can produce informative responses with real-world visual concepts. Our code is available at https://github.com/ImKeTT/ReSee.", }

Incorporating visual knowledge into text-only dialogue systems has become a potential direction to imitate the way humans think, imagine, and communicate. However, existing multimodal dialogue systems are either confined by the scale and quality of available datasets or the coarse concept of visual knowledge. To address these issues, we provide a new paradigm of constructing multimodal dialogues as well as two datasets extended from text-only dialogues under such paradigm (ReSee-$\texttt{WoW}$, ReSee-$\texttt{DD}$). We propose to explicitly split the visual knowledge into finer granularity ({``}turn-level{''} and {``}entity-level{''}). To further boost the accuracy and diversity of augmented visual information, we retrieve them from the Internet or a large image dataset. To demonstrate the superiority and universality of the provided visual knowledge, we propose a simple but effective framework ReSee to add visual representation into vanilla dialogue models by modality concatenations. We also conduct extensive experiments and ablations w.r.t. different model configurations and visual knowledge settings. Empirical, encouraging results not only demonstrate the effectiveness of introducing visual knowledge at both entity and turn level but also verify the proposed model ReSee outperforms several state-of-the-art methods on automatic and human evaluations. By leveraging text and vision knowledge, ReSee can produce informative responses with real-world visual concepts. Our code is available at https://github.com/ImKeTT/ReSee.

ReSee: Responding through Seeing Fine-grained Visual Knowledge in Open-domain Dialogue

emnlp-main.479

2305.13602

https://huggingface.co/papers/2305.13602

Oral

https://aclanthology.org/2023.emnlp-main.480.bib

https://aclanthology.org/2023.emnlp-main.480/

@inproceedings{tian-etal-2023-modeling, title = "Modeling Conceptual Attribute Likeness and Domain Inconsistency for Metaphor Detection", author = "Tian, Yuan and Xu, Nan and Mao, Wenji and Zeng, Daniel", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.480", doi = "10.18653/v1/2023.emnlp-main.480", pages = "7736--7752", abstract = "Metaphor detection is an important and challenging task in natural language processing, which aims to distinguish between metaphorical and literal expressions in text. Previous studies mainly leverage the incongruity of source and target domains and contextual clues for detection, neglecting similar attributes shared between source and target concepts in metaphorical expressions. Based on conceptual metaphor theory, these similar attributes are essential to infer implicit meanings conveyed by the metaphor. Under the guidance of conceptual metaphor theory, in this paper, we model the likeness of attribute for the first time and propose a novel Attribute Likeness and Domain Inconsistency Learning framework (AIDIL) for word-pair metaphor detection. Specifically, we propose an attribute siamese network to mine similar attributes between source and target concepts. We then devise a domain contrastive learning strategy to learn the semantic inconsistency of concepts in source and target domains. Extensive experiments on four datasets verify that our method significantly outperforms the previous state-of-the-art methods, and demonstrate the generalization ability of our method.", }

Metaphor detection is an important and challenging task in natural language processing, which aims to distinguish between metaphorical and literal expressions in text. Previous studies mainly leverage the incongruity of source and target domains and contextual clues for detection, neglecting similar attributes shared between source and target concepts in metaphorical expressions. Based on conceptual metaphor theory, these similar attributes are essential to infer implicit meanings conveyed by the metaphor. Under the guidance of conceptual metaphor theory, in this paper, we model the likeness of attribute for the first time and propose a novel Attribute Likeness and Domain Inconsistency Learning framework (AIDIL) for word-pair metaphor detection. Specifically, we propose an attribute siamese network to mine similar attributes between source and target concepts. We then devise a domain contrastive learning strategy to learn the semantic inconsistency of concepts in source and target domains. Extensive experiments on four datasets verify that our method significantly outperforms the previous state-of-the-art methods, and demonstrate the generalization ability of our method.

Modeling Conceptual Attribute Likeness and Domain Inconsistency for Metaphor Detection

emnlp-main.480

Poster

https://aclanthology.org/2023.emnlp-main.481.bib

https://aclanthology.org/2023.emnlp-main.481/

@inproceedings{huang-satoh-2023-referring, title = "Referring Image Segmentation via Joint Mask Contextual Embedding Learning and Progressive Alignment Network", author = "Huang, Ziling and Satoh, Shin{'}ichi", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.481", doi = "10.18653/v1/2023.emnlp-main.481", pages = "7753--7762", abstract = "Referring image segmentation is a task that aims to predict pixel-wise masks corresponding to objects in an image described by natural language expressions. Previous methods for referring image segmentation employ a cascade framework to break down complex problems into multiple stages. However, its defects also obvious: existing methods within the cascade framework may encounter challenges in both maintaining a strong focus on the most relevant information during specific stages of the referring image segmentation process and rectifying errors propagated from early stages, which can ultimately result in sub-optimal performance. To address these limitations, we propose the Joint Mask Contextual Embedding Learning Network (JMCELN). JMCELN is designed to enhance the Cascade Framework by incorporating a Learnable Contextual Embedding and a Progressive Alignment Network (PAN). The Learnable Contextual Embedding module dynamically stores and utilizes reasoning information based on the current mask prediction results, enabling the network to adaptively capture and refine pertinent information for improved mask prediction accuracy. Furthermore, the Progressive Alignment Network (PAN) is introduced as an integral part of JMCELN. PAN leverages the output from the previous layer as a filter for the current output, effectively reducing inconsistencies between predictions from different stages. By iteratively aligning the predictions, PAN guides the Learnable Contextual Embedding to incorporate more discriminative information for reasoning, leading to enhanced prediction quality and a reduction in error propagation. With these methods, we achieved state-of-the-art results on three commonly used benchmarks, especially in more intricate datasets. The code will be released.", }

Referring image segmentation is a task that aims to predict pixel-wise masks corresponding to objects in an image described by natural language expressions. Previous methods for referring image segmentation employ a cascade framework to break down complex problems into multiple stages. However, its defects also obvious: existing methods within the cascade framework may encounter challenges in both maintaining a strong focus on the most relevant information during specific stages of the referring image segmentation process and rectifying errors propagated from early stages, which can ultimately result in sub-optimal performance. To address these limitations, we propose the Joint Mask Contextual Embedding Learning Network (JMCELN). JMCELN is designed to enhance the Cascade Framework by incorporating a Learnable Contextual Embedding and a Progressive Alignment Network (PAN). The Learnable Contextual Embedding module dynamically stores and utilizes reasoning information based on the current mask prediction results, enabling the network to adaptively capture and refine pertinent information for improved mask prediction accuracy. Furthermore, the Progressive Alignment Network (PAN) is introduced as an integral part of JMCELN. PAN leverages the output from the previous layer as a filter for the current output, effectively reducing inconsistencies between predictions from different stages. By iteratively aligning the predictions, PAN guides the Learnable Contextual Embedding to incorporate more discriminative information for reasoning, leading to enhanced prediction quality and a reduction in error propagation. With these methods, we achieved state-of-the-art results on three commonly used benchmarks, especially in more intricate datasets. The code will be released.

Referring Image Segmentation via Joint Mask Contextual Embedding Learning and Progressive Alignment Network

emnlp-main.481

Poster

https://aclanthology.org/2023.emnlp-main.482.bib

https://aclanthology.org/2023.emnlp-main.482/

@inproceedings{wang-etal-2023-shall, title = "Shall We Pretrain Autoregressive Language Models with Retrieval? A Comprehensive Study", author = "Wang, Boxin and Ping, Wei and Xu, Peng and McAfee, Lawrence and Liu, Zihan and Shoeybi, Mohammad and Dong, Yi and Kuchaiev, Oleksii and Li, Bo and Xiao, Chaowei and Anandkumar, Anima and Catanzaro, Bryan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.482", doi = "10.18653/v1/2023.emnlp-main.482", pages = "7763--7786", abstract = "Large decoder-only language models (LMs) can be largely improved in terms of perplexity by retrieval (e.g., RETRO), but its impact on text generation quality and downstream task accuracy is unclear. Thus, it is still an open question: shall we pretrain large autoregressive LMs with retrieval? To answer it, we perform a comprehensive study on a scalable pre-trained retrieval-augmented LM (i.e., RETRO) compared with standard GPT and retrieval-augmented GPT incorporated at fine-tuning or inference stages. We first provide the recipe to reproduce RETRO up to 9.5B parameters while retrieving a text corpus with 330B tokens. Based on that, we have the following novel findings: i) RETRO outperforms GPT on text generation with much less degeneration (i.e., repetition), moderately higher factual accuracy, and slightly lower toxicity with a nontoxic retrieval database. ii) On the LM Evaluation Harness benchmark, RETRO largely outperforms GPT on knowledge-intensive tasks, but is on par with GPT on other tasks. Furthermore, we introduce a simple variant of the model, RETRO++, which largely improves open-domain QA results of original RETRO (e.g., EM score +8.6 on Natural Question) and significantly outperforms retrieval-augmented GPT across different model sizes. Our findings highlight the promising direction of pretraining autoregressive LMs with retrieval as future foundation models. We release our implementation at: https://github.com/NVIDIA/Megatron-LM/tree/main/tools/retro.", }

Large decoder-only language models (LMs) can be largely improved in terms of perplexity by retrieval (e.g., RETRO), but its impact on text generation quality and downstream task accuracy is unclear. Thus, it is still an open question: shall we pretrain large autoregressive LMs with retrieval? To answer it, we perform a comprehensive study on a scalable pre-trained retrieval-augmented LM (i.e., RETRO) compared with standard GPT and retrieval-augmented GPT incorporated at fine-tuning or inference stages. We first provide the recipe to reproduce RETRO up to 9.5B parameters while retrieving a text corpus with 330B tokens. Based on that, we have the following novel findings: i) RETRO outperforms GPT on text generation with much less degeneration (i.e., repetition), moderately higher factual accuracy, and slightly lower toxicity with a nontoxic retrieval database. ii) On the LM Evaluation Harness benchmark, RETRO largely outperforms GPT on knowledge-intensive tasks, but is on par with GPT on other tasks. Furthermore, we introduce a simple variant of the model, RETRO++, which largely improves open-domain QA results of original RETRO (e.g., EM score +8.6 on Natural Question) and significantly outperforms retrieval-augmented GPT across different model sizes. Our findings highlight the promising direction of pretraining autoregressive LMs with retrieval as future foundation models. We release our implementation at: https://github.com/NVIDIA/Megatron-LM/tree/main/tools/retro.

Shall We Pretrain Autoregressive Language Models with Retrieval? A Comprehensive Study

emnlp-main.482

2304.06762

https://huggingface.co/papers/2304.06762

Poster

https://aclanthology.org/2023.emnlp-main.483.bib

https://aclanthology.org/2023.emnlp-main.483/

@inproceedings{lu-etal-2023-scitab, title = "{SCITAB}: A Challenging Benchmark for Compositional Reasoning and Claim Verification on Scientific Tables", author = "Lu, Xinyuan and Pan, Liangming and Liu, Qian and Nakov, Preslav and Kan, Min-Yen", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.483", doi = "10.18653/v1/2023.emnlp-main.483", pages = "7787--7813", abstract = "Current scientific fact-checking benchmarks exhibit several shortcomings, such as biases arising from crowd-sourced claims and an over-reliance on text-based evidence. We present SCITAB, a challenging evaluation dataset consisting of 1.2K expert-verified scientific claims that 1) originate from authentic scientific publications and 2) require compositional reasoning for verification. The claims are paired with evidence-containing scientific tables annotated with labels. Through extensive evaluations, we demonstrate that SCITAB poses a significant challenge to state-of-the-art models, including table-based pretraining models and large language models. All models except GPT-4 achieved performance barely above random guessing. Popular prompting techniques, such as Chain-of-Thought, do not achieve much performance gains on SCITAB. Our analysis uncovers several unique challenges posed by SCITAB, including table grounding, claim ambiguity, and compositional reasoning. Our codes and data are publicly available at https://github.com/XinyuanLu00/SciTab.", }

Current scientific fact-checking benchmarks exhibit several shortcomings, such as biases arising from crowd-sourced claims and an over-reliance on text-based evidence. We present SCITAB, a challenging evaluation dataset consisting of 1.2K expert-verified scientific claims that 1) originate from authentic scientific publications and 2) require compositional reasoning for verification. The claims are paired with evidence-containing scientific tables annotated with labels. Through extensive evaluations, we demonstrate that SCITAB poses a significant challenge to state-of-the-art models, including table-based pretraining models and large language models. All models except GPT-4 achieved performance barely above random guessing. Popular prompting techniques, such as Chain-of-Thought, do not achieve much performance gains on SCITAB. Our analysis uncovers several unique challenges posed by SCITAB, including table grounding, claim ambiguity, and compositional reasoning. Our codes and data are publicly available at https://github.com/XinyuanLu00/SciTab.

SCITAB: A Challenging Benchmark for Compositional Reasoning and Claim Verification on Scientific Tables

emnlp-main.483

2305.13186

Poster

https://aclanthology.org/2023.emnlp-main.484.bib

https://aclanthology.org/2023.emnlp-main.484/

@inproceedings{zhang-etal-2023-training, title = "Training Simultaneous Speech Translation with Robust and Random Wait-k-Tokens Strategy", author = "Zhang, Linlin and Fan, Kai and Bu, Jiajun and Huang, Zhongqiang", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.484", doi = "10.18653/v1/2023.emnlp-main.484", pages = "7814--7831", abstract = "Simultaneous Speech Translation (SimulST) is a task focused on ensuring high-quality translation of speech in low-latency situations. Despite this, the modality gap (\textit{e.g.}, unknown word boundaries) between audio and text presents a challenge. This gap hinders the effective application of policies from simultaneous text translation (SimulMT) and compromises the performance of offline speech translation. To address this issue, we first leverage the Montreal Forced Aligner (MFA) and utilize audio transcription pairs in pre-training the acoustic encoder, and introduce a token-level cross-modal alignment that allows the wait-$k$ policy from SimulMT to better adapt to SimulST. This token-level boundary alignment simplifies the decision-making process for predicting read/write actions, as if the decoder were directly processing text tokens. Subsequently, to optimize the SimulST task, we propose a robust and random wait-$k$-tokens strategy. This strategy allows a single model to meet various latency requirements and minimizes error accumulation of boundary alignment during inference. Our experiments on the MuST-C dataset show that our method achieves better trade-off between translation quality and latency.", }

Simultaneous Speech Translation (SimulST) is a task focused on ensuring high-quality translation of speech in low-latency situations. Despite this, the modality gap (\textit{e.g.}, unknown word boundaries) between audio and text presents a challenge. This gap hinders the effective application of policies from simultaneous text translation (SimulMT) and compromises the performance of offline speech translation. To address this issue, we first leverage the Montreal Forced Aligner (MFA) and utilize audio transcription pairs in pre-training the acoustic encoder, and introduce a token-level cross-modal alignment that allows the wait-$k$ policy from SimulMT to better adapt to SimulST. This token-level boundary alignment simplifies the decision-making process for predicting read/write actions, as if the decoder were directly processing text tokens. Subsequently, to optimize the SimulST task, we propose a robust and random wait-$k$-tokens strategy. This strategy allows a single model to meet various latency requirements and minimizes error accumulation of boundary alignment during inference. Our experiments on the MuST-C dataset show that our method achieves better trade-off between translation quality and latency.

Training Simultaneous Speech Translation with Robust and Random Wait-k-Tokens Strategy

emnlp-main.484

Poster

https://aclanthology.org/2023.emnlp-main.485.bib

https://aclanthology.org/2023.emnlp-main.485/

@inproceedings{fu-etal-2023-scene, title = "{SCENE}: Self-Labeled Counterfactuals for Extrapolating to Negative Examples", author = "Fu, Deqing and Godbole, Ameya and Jia, Robin", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.485", doi = "10.18653/v1/2023.emnlp-main.485", pages = "7832--7848", abstract = "Detecting negatives (such as non-entailment relationships, unanswerable questions, and false claims) is an important and challenging aspect of many natural language understanding tasks. Though manually collecting challenging negative examples can help models detect them, it is both costly and domain-specific. In this work, we propose Self-labeled Counterfactuals for Extrapolating to Negative Examples (SCENE), an automatic method for synthesizing training data that greatly improves models{'} ability to detect challenging negative examples. In contrast with standard data augmentation, which synthesizes new examples for existing labels, SCENE can synthesize negative examples zero-shot from only positive ones. Given a positive example, SCENE perturbs it with a mask infilling model, then determines whether the resulting example is negative based on a self-training heuristic. With access to only answerable training examples, SCENE can close 69.6{\%} of the performance gap on SQuAD 2.0, a dataset where half of the evaluation examples are unanswerable, compared to a model trained on SQuAD 2.0. Our method also extends to boolean question answering and recognizing textual entailment, and improves generalization from SQuAD to ACE-whQA, an out-of-domain extractive QA benchmark.", }

Detecting negatives (such as non-entailment relationships, unanswerable questions, and false claims) is an important and challenging aspect of many natural language understanding tasks. Though manually collecting challenging negative examples can help models detect them, it is both costly and domain-specific. In this work, we propose Self-labeled Counterfactuals for Extrapolating to Negative Examples (SCENE), an automatic method for synthesizing training data that greatly improves models{'} ability to detect challenging negative examples. In contrast with standard data augmentation, which synthesizes new examples for existing labels, SCENE can synthesize negative examples zero-shot from only positive ones. Given a positive example, SCENE perturbs it with a mask infilling model, then determines whether the resulting example is negative based on a self-training heuristic. With access to only answerable training examples, SCENE can close 69.6{\%} of the performance gap on SQuAD 2.0, a dataset where half of the evaluation examples are unanswerable, compared to a model trained on SQuAD 2.0. Our method also extends to boolean question answering and recognizing textual entailment, and improves generalization from SQuAD to ACE-whQA, an out-of-domain extractive QA benchmark.

SCENE: Self-Labeled Counterfactuals for Extrapolating to Negative Examples

emnlp-main.485

2305.07984

Poster

https://aclanthology.org/2023.emnlp-main.486.bib

https://aclanthology.org/2023.emnlp-main.486/

@inproceedings{zhu-etal-2023-enhancing, title = "Enhancing Code-Switching for Cross-lingual {SLU}: A Unified View of Semantic and Grammatical Coherence", author = "Zhu, Zhihong and Cheng, Xuxin and Huang, Zhiqi and Chen, Dongsheng and Zou, Yuexian", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.486", doi = "10.18653/v1/2023.emnlp-main.486", pages = "7849--7856", abstract = "Despite the success of spoken language understanding (SLU) in high-resource languages, achieving similar performance in low-resource settings, such as zero-shot scenarios, remains challenging due to limited labeled training data. To improve zero-shot cross-lingual SLU, recent studies have explored code-switched sentences containing tokens from multiple languages. However, vanilla code-switched sentences often lack semantic and grammatical coherence. We ascribe this lack to two issues: (1) randomly replacing code-switched tokens with equal probability and (2) disregarding token-level dependency within each language. To tackle these issues, in this paper, we propose a novel method termed SoGo, for zero-shot cross-lingual SLU. First, we use a saliency-based substitution approach to extract keywords as substitution options. Then, we introduce a novel token-level alignment strategy that considers the similarity between the context and the code-switched tokens, ensuring grammatical coherence in code-switched sentences. Extensive experiments and analyses demonstrate the superior performance of SoGo across nine languages on MultiATIS++.", }

Despite the success of spoken language understanding (SLU) in high-resource languages, achieving similar performance in low-resource settings, such as zero-shot scenarios, remains challenging due to limited labeled training data. To improve zero-shot cross-lingual SLU, recent studies have explored code-switched sentences containing tokens from multiple languages. However, vanilla code-switched sentences often lack semantic and grammatical coherence. We ascribe this lack to two issues: (1) randomly replacing code-switched tokens with equal probability and (2) disregarding token-level dependency within each language. To tackle these issues, in this paper, we propose a novel method termed SoGo, for zero-shot cross-lingual SLU. First, we use a saliency-based substitution approach to extract keywords as substitution options. Then, we introduce a novel token-level alignment strategy that considers the similarity between the context and the code-switched tokens, ensuring grammatical coherence in code-switched sentences. Extensive experiments and analyses demonstrate the superior performance of SoGo across nine languages on MultiATIS++.

Enhancing Code-Switching for Cross-lingual SLU: A Unified View of Semantic and Grammatical Coherence

emnlp-main.486

Oral

https://aclanthology.org/2023.emnlp-main.487.bib

https://aclanthology.org/2023.emnlp-main.487/

@inproceedings{xiao-etal-2023-task, title = "Task-Agnostic Low-Rank Adapters for Unseen {E}nglish Dialects", author = "Xiao, Zedian and Held, William and Liu, Yanchen and Yang, Diyi", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.487", doi = "10.18653/v1/2023.emnlp-main.487", pages = "7857--7870", abstract = "Large Language Models (LLMs) are trained on corpora disproportionally weighted in favor of Standard American English. As a result, speakers of other dialects experience significantly more failures when interacting with these technologies. In practice, these speakers often accommodate their speech to be better understood. Our work shares the belief that language technologies should be designed to accommodate the diversity in English dialects and not the other way around. However, prior work on dialect struggle with generalizing to evolving and emerging dialects in a scalable manner. To fill this gap, our method, HyperLoRA, leverages expert linguistic knowledge to enable resource-efficient adaptation via hypernetworks. By disentangling dialect-specific and cross-dialectal information, HyperLoRA improves generalization to unseen dialects in a task-agnostic fashion. Not only is HyperLoRA more scalable in the number of parameters, but it also achieves the best or most competitive performance across 5 dialects in a zero-shot setting. In this way, our approach facilitates access to language technology for billions of English dialect speakers who are traditionally underrepresented.", }

Large Language Models (LLMs) are trained on corpora disproportionally weighted in favor of Standard American English. As a result, speakers of other dialects experience significantly more failures when interacting with these technologies. In practice, these speakers often accommodate their speech to be better understood. Our work shares the belief that language technologies should be designed to accommodate the diversity in English dialects and not the other way around. However, prior work on dialect struggle with generalizing to evolving and emerging dialects in a scalable manner. To fill this gap, our method, HyperLoRA, leverages expert linguistic knowledge to enable resource-efficient adaptation via hypernetworks. By disentangling dialect-specific and cross-dialectal information, HyperLoRA improves generalization to unseen dialects in a task-agnostic fashion. Not only is HyperLoRA more scalable in the number of parameters, but it also achieves the best or most competitive performance across 5 dialects in a zero-shot setting. In this way, our approach facilitates access to language technology for billions of English dialect speakers who are traditionally underrepresented.

Task-Agnostic Low-Rank Adapters for Unseen English Dialects

emnlp-main.487

2311.00915

https://huggingface.co/papers/2311.00915

Poster

https://aclanthology.org/2023.emnlp-main.488.bib

https://aclanthology.org/2023.emnlp-main.488/

@inproceedings{che-etal-2023-federated, title = "Federated Learning of Large Language Models with Parameter-Efficient Prompt Tuning and Adaptive Optimization", author = "Che, Tianshi and Liu, Ji and Zhou, Yang and Ren, Jiaxiang and Zhou, Jiwen and Sheng, Victor and Dai, Huaiyu and Dou, Dejing", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.488", doi = "10.18653/v1/2023.emnlp-main.488", pages = "7871--7888", abstract = "Federated learning (FL) is a promising paradigm to enable collaborative model training with decentralized data. However, the training process of Large Language Models (LLMs) generally incurs the update of significant parameters, which limits the applicability of FL techniques to tackle the LLMs in real scenarios. Prompt tuning can significantly reduce the number of parameters to update, but it either incurs performance degradation or low training efficiency. The straightforward utilization of prompt tuning in the FL often raises non-trivial communication costs and dramatically degrades performance. In addition, the decentralized data is generally non-Independent and Identically Distributed (non-IID), which brings client drift problems and thus poor performance. This paper proposes a Parameter-efficient prompt Tuning approach with Adaptive Optimization, i.e., FedPepTAO, to enable efficient and effective FL of LLMs. First, an efficient partial prompt tuning approach is proposed to improve performance and efficiency simultaneously. Second, a novel adaptive optimization method is developed to address the client drift problems on both the device and server sides to enhance performance further. Extensive experiments based on 10 datasets demonstrate the superb performance (up to 60.8{\%} in terms of accuracy) and efficiency (up to 97.59{\%} in terms of training time) of FedPepTAO compared with 9 baseline approaches. Our code is available at https://github.com/llm-eff/FedPepTAO.", }

Federated learning (FL) is a promising paradigm to enable collaborative model training with decentralized data. However, the training process of Large Language Models (LLMs) generally incurs the update of significant parameters, which limits the applicability of FL techniques to tackle the LLMs in real scenarios. Prompt tuning can significantly reduce the number of parameters to update, but it either incurs performance degradation or low training efficiency. The straightforward utilization of prompt tuning in the FL often raises non-trivial communication costs and dramatically degrades performance. In addition, the decentralized data is generally non-Independent and Identically Distributed (non-IID), which brings client drift problems and thus poor performance. This paper proposes a Parameter-efficient prompt Tuning approach with Adaptive Optimization, i.e., FedPepTAO, to enable efficient and effective FL of LLMs. First, an efficient partial prompt tuning approach is proposed to improve performance and efficiency simultaneously. Second, a novel adaptive optimization method is developed to address the client drift problems on both the device and server sides to enhance performance further. Extensive experiments based on 10 datasets demonstrate the superb performance (up to 60.8{\%} in terms of accuracy) and efficiency (up to 97.59{\%} in terms of training time) of FedPepTAO compared with 9 baseline approaches. Our code is available at https://github.com/llm-eff/FedPepTAO.

Federated Learning of Large Language Models with Parameter-Efficient Prompt Tuning and Adaptive Optimization

emnlp-main.488

2310.15080

Poster

https://aclanthology.org/2023.emnlp-main.489.bib

https://aclanthology.org/2023.emnlp-main.489/

@inproceedings{chen-etal-2023-theoremqa, title = "{T}heorem{QA}: A Theorem-driven Question Answering Dataset", author = "Chen, Wenhu and Yin, Ming and Ku, Max and Lu, Pan and Wan, Yixin and Ma, Xueguang and Xu, Jianyu and Wang, Xinyi and Xia, Tony", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.489", doi = "10.18653/v1/2023.emnlp-main.489", pages = "7889--7901", abstract = "The recent LLMs like GPT-4 and PaLM-2 have made tremendous progress in solving fundamental math problems like GSM8K by achieving over 90{\%} accuracy. However, their capabilities to solve more challenging math problems which require domain-specific knowledge (i.e. theorem) have yet to be investigated. In this paper, we introduce TheoremQA, the first theorem-driven question-answering dataset designed to evaluate AI models{'} capabilities to apply theorems to solve challenging science problems. TheoremQA is curated by domain experts containing 800 high-quality questions covering 350 theorems from Math, Physics, EE{\&}CS, and Finance. We evaluate a wide spectrum of 16 large language and code models with different prompting strategies like Chain-of-Thoughts and Program-of-Thoughts. We found that GPT-4{'}s capabilities to solve these problems are unparalleled, achieving an accuracy of 51{\%} with Program-of-Thoughts Prompting. All the existing open-sourced models are below 15{\%}, barely surpassing the random-guess baseline. Given the diversity and broad coverage of TheoremQA, we believe it can be used as a better benchmark to evaluate LLMs{'} capabilities to solve challenging science problems.", }

The recent LLMs like GPT-4 and PaLM-2 have made tremendous progress in solving fundamental math problems like GSM8K by achieving over 90{\%} accuracy. However, their capabilities to solve more challenging math problems which require domain-specific knowledge (i.e. theorem) have yet to be investigated. In this paper, we introduce TheoremQA, the first theorem-driven question-answering dataset designed to evaluate AI models{'} capabilities to apply theorems to solve challenging science problems. TheoremQA is curated by domain experts containing 800 high-quality questions covering 350 theorems from Math, Physics, EE{\&}CS, and Finance. We evaluate a wide spectrum of 16 large language and code models with different prompting strategies like Chain-of-Thoughts and Program-of-Thoughts. We found that GPT-4{'}s capabilities to solve these problems are unparalleled, achieving an accuracy of 51{\%} with Program-of-Thoughts Prompting. All the existing open-sourced models are below 15{\%}, barely surpassing the random-guess baseline. Given the diversity and broad coverage of TheoremQA, we believe it can be used as a better benchmark to evaluate LLMs{'} capabilities to solve challenging science problems.

TheoremQA: A Theorem-driven Question Answering Dataset

emnlp-main.489

2305.12524

https://huggingface.co/papers/2305.12524

Poster

https://aclanthology.org/2023.emnlp-main.490.bib

https://aclanthology.org/2023.emnlp-main.490/

@inproceedings{su-etal-2023-scalable, title = "Scalable-{DSC}: A Structural Template Prompt Approach to Scalable Dialogue State Correction", author = "Su, Haoxiang and Xie, Hongyan and Huang, Hao and Song, Shuangyong and Fang, Ruiyu and Huang, Xiaomeng and Feng, Sijie", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.490", doi = "10.18653/v1/2023.emnlp-main.490", pages = "7902--7914", abstract = "Dialogue state error correction has recently been proposed to correct wrong slot values in predicted dialogue states, thereby mitigating the error propagation problem for dialogue state tracking (DST). These approaches, though effective, are heavily intertwined with specific DST models, limiting their applicability to other DST models. To solve this problem, we propose Scalable Dialogue State Correction (Scalable-DSC), which can correct wrong slot values in the dialogue state predicted by any DST model. Specifically, we propose a Structural Template Prompt (STP) that converts predicted dialogue state from any DST models into a standardized natural language sequence as a part of the historical context, associates them with dialogue history information, and generates a corrected dialogue state sequence based on predefined template options. We further enhance Scalable-DSC by introducing two training strategies. The first employs a predictive state simulator to simulate the predicted dialogue states as the training data to enhance the generalization ability of the model. The second involves using the dialogue state predicted by DST as the training data, aiming at mitigating the inconsistent error type distribution between the training and inference. Experiments confirm that our model achieves state-of-the-art results on MultiWOZ 2.0-2.4.", }

Dialogue state error correction has recently been proposed to correct wrong slot values in predicted dialogue states, thereby mitigating the error propagation problem for dialogue state tracking (DST). These approaches, though effective, are heavily intertwined with specific DST models, limiting their applicability to other DST models. To solve this problem, we propose Scalable Dialogue State Correction (Scalable-DSC), which can correct wrong slot values in the dialogue state predicted by any DST model. Specifically, we propose a Structural Template Prompt (STP) that converts predicted dialogue state from any DST models into a standardized natural language sequence as a part of the historical context, associates them with dialogue history information, and generates a corrected dialogue state sequence based on predefined template options. We further enhance Scalable-DSC by introducing two training strategies. The first employs a predictive state simulator to simulate the predicted dialogue states as the training data to enhance the generalization ability of the model. The second involves using the dialogue state predicted by DST as the training data, aiming at mitigating the inconsistent error type distribution between the training and inference. Experiments confirm that our model achieves state-of-the-art results on MultiWOZ 2.0-2.4.

Scalable-DSC: A Structural Template Prompt Approach to Scalable Dialogue State Correction

emnlp-main.490

Poster

https://aclanthology.org/2023.emnlp-main.491.bib

https://aclanthology.org/2023.emnlp-main.491/

@inproceedings{zhang-etal-2023-dont, title = "Don{'}t Trust {C}hat{GPT} when your Question is not in {E}nglish: A Study of Multilingual Abilities and Types of {LLM}s", author = "Zhang, Xiang and Li, Senyu and Hauer, Bradley and Shi, Ning and Kondrak, Grzegorz", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.491", doi = "10.18653/v1/2023.emnlp-main.491", pages = "7915--7927", abstract = "Large language models (LLMs) have demonstrated exceptional natural language understanding abilities, and have excelled in a variety of natural language processing (NLP) tasks. Despite the fact that most LLMs are trained predominantly on English, multiple studies have demonstrated their capabilities in a variety of languages. However, fundamental questions persist regarding how LLMs acquire their multilingual abilities and how performance varies across different languages. These inquiries are crucial for the study of LLMs since users and researchers often come from diverse language backgrounds, potentially influencing how they use LLMs and interpret their output. In this work, we propose a systematic way of qualitatively and quantitatively evaluating the multilingual capabilities of LLMs. We investigate the phenomenon of cross-language generalization in LLMs, wherein limited multilingual training data leads to advanced multilingual capabilities. To accomplish this, we employ a novel prompt back-translation method. The results demonstrate that LLMs, such as GPT, can effectively transfer learned knowledge across different languages, yielding relatively consistent results in translation-equivariant tasks, in which the correct output does not depend on the language of the input. However, LLMs struggle to provide accurate results in translation-variant tasks, which lack this property, requiring careful user judgment to evaluate the answers.", }

Large language models (LLMs) have demonstrated exceptional natural language understanding abilities, and have excelled in a variety of natural language processing (NLP) tasks. Despite the fact that most LLMs are trained predominantly on English, multiple studies have demonstrated their capabilities in a variety of languages. However, fundamental questions persist regarding how LLMs acquire their multilingual abilities and how performance varies across different languages. These inquiries are crucial for the study of LLMs since users and researchers often come from diverse language backgrounds, potentially influencing how they use LLMs and interpret their output. In this work, we propose a systematic way of qualitatively and quantitatively evaluating the multilingual capabilities of LLMs. We investigate the phenomenon of cross-language generalization in LLMs, wherein limited multilingual training data leads to advanced multilingual capabilities. To accomplish this, we employ a novel prompt back-translation method. The results demonstrate that LLMs, such as GPT, can effectively transfer learned knowledge across different languages, yielding relatively consistent results in translation-equivariant tasks, in which the correct output does not depend on the language of the input. However, LLMs struggle to provide accurate results in translation-variant tasks, which lack this property, requiring careful user judgment to evaluate the answers.

Don't Trust ChatGPT when your Question is not in English: A Study of Multilingual Abilities and Types of LLMs

emnlp-main.491

2305.16339

Oral

https://aclanthology.org/2023.emnlp-main.492.bib

https://aclanthology.org/2023.emnlp-main.492/

@inproceedings{wang-etal-2023-m3seg, title = "{M}$^3${S}eg: A Maximum-Minimum Mutual Information Paradigm for Unsupervised Topic Segmentation in {ASR} Transcripts", author = "Wang, Ke and Zhao, Xiutian and Li, Yanghui and Peng, Wei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.492", doi = "10.18653/v1/2023.emnlp-main.492", pages = "7928--7934", abstract = "Topic segmentation aims to detect topic boundaries and split automatic speech recognition transcriptions (e.g., meeting transcripts) into segments that are bounded by thematic meanings. In this work, we propose M$^3$Seg, a novel Maximum-Minimum Mutual information paradigm for linear topic segmentation without using any parallel data. Specifically, by employing sentence representations provided by pre-trained language models, M$^3$Seg first learns a region-based segment encoder based on the maximization of mutual information between the global segment representation and the local contextual sentence representation. Secondly, an edge-based boundary detection module aims to segment the whole by topics based on minimizing the mutual information between different segments. Experiment results on two public datasets demonstrate the effectiveness of M$^3$Seg, which outperform the state-of-the-art methods by a significant (18{\%}{--}37{\%} improvement) margin.", }

Topic segmentation aims to detect topic boundaries and split automatic speech recognition transcriptions (e.g., meeting transcripts) into segments that are bounded by thematic meanings. In this work, we propose M$^3$Seg, a novel Maximum-Minimum Mutual information paradigm for linear topic segmentation without using any parallel data. Specifically, by employing sentence representations provided by pre-trained language models, M$^3$Seg first learns a region-based segment encoder based on the maximization of mutual information between the global segment representation and the local contextual sentence representation. Secondly, an edge-based boundary detection module aims to segment the whole by topics based on minimizing the mutual information between different segments. Experiment results on two public datasets demonstrate the effectiveness of M$^3$Seg, which outperform the state-of-the-art methods by a significant (18{\%}{--}37{\%} improvement) margin.

M^3Seg: A Maximum-Minimum Mutual Information Paradigm for Unsupervised Topic Segmentation in ASR Transcripts

emnlp-main.492

Poster

https://aclanthology.org/2023.emnlp-main.493.bib

https://aclanthology.org/2023.emnlp-main.493/

@inproceedings{xie-etal-2023-empirical, title = "Empirical Study of Zero-Shot {NER} with {C}hat{GPT}", author = "Xie, Tingyu and Li, Qi and Zhang, Jian and Zhang, Yan and Liu, Zuozhu and Wang, Hongwei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.493", doi = "10.18653/v1/2023.emnlp-main.493", pages = "7935--7956", abstract = "Large language models (LLMs) exhibited powerful capability in various natural language processing tasks. This work focuses on exploring LLM performance on zero-shot information extraction, with a focus on the ChatGPT and named entity recognition (NER) task. Inspired by the remarkable reasoning capability of LLM on symbolic and arithmetic reasoning, we adapt the prevalent reasoning methods to NER and propose reasoning strategies tailored for NER. First, we explore a decomposed question-answering paradigm by breaking down the NER task into simpler subproblems by labels. Second, we propose syntactic augmentation to stimulate the model{'}s intermediate thinking in two ways: syntactic prompting, which encourages the model to analyze the syntactic structure itself, and tool augmentation, which provides the model with the syntactic information generated by a parsing tool. Besides, we adapt self-consistency to NER by proposing a two-stage majority voting strategy, which first votes for the most consistent mentions, then the most consistent types. The proposed methods achieve remarkable improvements for zero-shot NER across seven benchmarks, including Chinese and English datasets, and on both domain-specific and general-domain scenarios. In addition, we present a comprehensive analysis of the error types with suggestions for optimization directions. We also verify the effectiveness of the proposed methods on the few-shot setting and other LLMs.", }

Large language models (LLMs) exhibited powerful capability in various natural language processing tasks. This work focuses on exploring LLM performance on zero-shot information extraction, with a focus on the ChatGPT and named entity recognition (NER) task. Inspired by the remarkable reasoning capability of LLM on symbolic and arithmetic reasoning, we adapt the prevalent reasoning methods to NER and propose reasoning strategies tailored for NER. First, we explore a decomposed question-answering paradigm by breaking down the NER task into simpler subproblems by labels. Second, we propose syntactic augmentation to stimulate the model{'}s intermediate thinking in two ways: syntactic prompting, which encourages the model to analyze the syntactic structure itself, and tool augmentation, which provides the model with the syntactic information generated by a parsing tool. Besides, we adapt self-consistency to NER by proposing a two-stage majority voting strategy, which first votes for the most consistent mentions, then the most consistent types. The proposed methods achieve remarkable improvements for zero-shot NER across seven benchmarks, including Chinese and English datasets, and on both domain-specific and general-domain scenarios. In addition, we present a comprehensive analysis of the error types with suggestions for optimization directions. We also verify the effectiveness of the proposed methods on the few-shot setting and other LLMs.

Empirical Study of Zero-Shot NER with ChatGPT

emnlp-main.493

2310.10035

Poster

https://aclanthology.org/2023.emnlp-main.494.bib

https://aclanthology.org/2023.emnlp-main.494/

@inproceedings{pryzant-etal-2023-automatic, title = "Automatic Prompt Optimization with {``}Gradient Descent{''} and Beam Search", author = "Pryzant, Reid and Iter, Dan and Li, Jerry and Lee, Yin and Zhu, Chenguang and Zeng, Michael", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.494", doi = "10.18653/v1/2023.emnlp-main.494", pages = "7957--7968", abstract = "Large Language Models (LLMs) have shown impressive performance as general purpose agents, but their abilities remain highly dependent on prompts which are hand written with onerous trial-and-error effort. We propose a simple and nonparametric solution to this problem, Prompt Optimization with Textual Gradients (ProTeGi), which is inspired by numerical gradient descent to automatically improve prompts, assuming access to training data and an LLM API. The algorithm uses minibatches of data to form natural language {``}gradients{''} that criticize the current prompt, much like how numerical gradients point in the direction of error ascent. The natural language gradients are then {``}propagated{''} into the prompt by editing the prompt in the opposite semantic direction of the gradient. These gradient descent steps are guided by a beam search and bandit selection procedure which significantly improves algorithmic efficiency. Preliminary results across three benchmark NLP tasks and the novel problem of LLM jailbreak detection suggest that Automatic Prompt Optimization can outperform prior prompt editing techniques and improve an initial prompt{'}s performance by up to 31{\%}, by using data to rewrite vague task descriptions into more precise annotation instructions.", }

Large Language Models (LLMs) have shown impressive performance as general purpose agents, but their abilities remain highly dependent on prompts which are hand written with onerous trial-and-error effort. We propose a simple and nonparametric solution to this problem, Prompt Optimization with Textual Gradients (ProTeGi), which is inspired by numerical gradient descent to automatically improve prompts, assuming access to training data and an LLM API. The algorithm uses minibatches of data to form natural language {``}gradients{''} that criticize the current prompt, much like how numerical gradients point in the direction of error ascent. The natural language gradients are then {``}propagated{''} into the prompt by editing the prompt in the opposite semantic direction of the gradient. These gradient descent steps are guided by a beam search and bandit selection procedure which significantly improves algorithmic efficiency. Preliminary results across three benchmark NLP tasks and the novel problem of LLM jailbreak detection suggest that Automatic Prompt Optimization can outperform prior prompt editing techniques and improve an initial prompt{'}s performance by up to 31{\%}, by using data to rewrite vague task descriptions into more precise annotation instructions.

Automatic Prompt Optimization with “Gradient Descent” and Beam Search

emnlp-main.494

2305.03495

https://huggingface.co/papers/2305.03495

Poster

https://aclanthology.org/2023.emnlp-main.495.bib

https://aclanthology.org/2023.emnlp-main.495/

@inproceedings{jiang-etal-2023-active, title = "Active Retrieval Augmented Generation", author = "Jiang, Zhengbao and Xu, Frank and Gao, Luyu and Sun, Zhiqing and Liu, Qian and Dwivedi-Yu, Jane and Yang, Yiming and Callan, Jamie and Neubig, Graham", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.495", doi = "10.18653/v1/2023.emnlp-main.495", pages = "7969--7992", abstract = "Despite the remarkable ability of large language models (LMs) to comprehend and generate language, they have a tendency to hallucinate and create factually inaccurate output. Augmenting LMs by retrieving information from external knowledge resources is one promising solution. Most existing retrieval augmented LMs employ a retrieve-and-generate setup that only retrieves information once based on the input. This is limiting, however, in more general scenarios involving generation of long texts, where continually gathering information throughout generation is essential. In this work, we provide a generalized view of active retrieval augmented generation, methods that actively decide when and what to retrieve across the course of the generation. We propose Forward-Looking Active REtrieval augmented generation (FLARE), a generic method which iteratively uses a prediction of the upcoming sentence to anticipate future content, which is then utilized as a query to retrieve relevant documents to regenerate the sentence if it contains low-confidence tokens. We test FLARE along with baselines comprehensively over 4 long-form knowledge-intensive generation tasks/datasets. FLARE achieves superior or competitive performance on all tasks, demonstrating the effectiveness of our method.", }

Despite the remarkable ability of large language models (LMs) to comprehend and generate language, they have a tendency to hallucinate and create factually inaccurate output. Augmenting LMs by retrieving information from external knowledge resources is one promising solution. Most existing retrieval augmented LMs employ a retrieve-and-generate setup that only retrieves information once based on the input. This is limiting, however, in more general scenarios involving generation of long texts, where continually gathering information throughout generation is essential. In this work, we provide a generalized view of active retrieval augmented generation, methods that actively decide when and what to retrieve across the course of the generation. We propose Forward-Looking Active REtrieval augmented generation (FLARE), a generic method which iteratively uses a prediction of the upcoming sentence to anticipate future content, which is then utilized as a query to retrieve relevant documents to regenerate the sentence if it contains low-confidence tokens. We test FLARE along with baselines comprehensively over 4 long-form knowledge-intensive generation tasks/datasets. FLARE achieves superior or competitive performance on all tasks, demonstrating the effectiveness of our method.

Active Retrieval Augmented Generation

emnlp-main.495

2406.12534

Poster

https://aclanthology.org/2023.emnlp-main.496.bib

https://aclanthology.org/2023.emnlp-main.496/

@inproceedings{bhatia-shwartz-2023-gd, title = "{GD}-{COMET}: A Geo-Diverse Commonsense Inference Model", author = "Bhatia, Mehar and Shwartz, Vered", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.496", doi = "10.18653/v1/2023.emnlp-main.496", pages = "7993--8001", abstract = "With the increasing integration of AI into everyday life, it{'}s becoming crucial to design AI systems to serve users from diverse backgrounds by making them culturally aware. In this paper, we present GD-COMET, a geo-diverse version of the COMET commonsense inference model. GD-COMET goes beyond Western commonsense knowledge and is capable of generating inferences pertaining to a broad range of cultures. We demonstrate the effectiveness of GD-COMET through a comprehensive human evaluation across 5 diverse cultures, as well as extrinsic evaluation on a geo-diverse task. The evaluation shows that GD-COMET captures and generates culturally nuanced commonsense knowledge, demonstrating its potential to benefit NLP applications across the board and contribute to making NLP more inclusive.", }

With the increasing integration of AI into everyday life, it{'}s becoming crucial to design AI systems to serve users from diverse backgrounds by making them culturally aware. In this paper, we present GD-COMET, a geo-diverse version of the COMET commonsense inference model. GD-COMET goes beyond Western commonsense knowledge and is capable of generating inferences pertaining to a broad range of cultures. We demonstrate the effectiveness of GD-COMET through a comprehensive human evaluation across 5 diverse cultures, as well as extrinsic evaluation on a geo-diverse task. The evaluation shows that GD-COMET captures and generates culturally nuanced commonsense knowledge, demonstrating its potential to benefit NLP applications across the board and contribute to making NLP more inclusive.

GD-COMET: A Geo-Diverse Commonsense Inference Model

emnlp-main.496

2310.15383

Oral

https://aclanthology.org/2023.emnlp-main.497.bib

https://aclanthology.org/2023.emnlp-main.497/

@inproceedings{yang-etal-2023-multi-level, title = "Multi-level Adaptive Contrastive Learning for Knowledge Internalization in Dialogue Generation", author = "Yang, Chenxu and Lin, Zheng and Wang, Lanrui and Tian, Chong and Pang, Liang and Li, Jiangnan and Ho, Qirong and Cao, Yanan and Wang, Weiping", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.497", doi = "10.18653/v1/2023.emnlp-main.497", pages = "8002--8015", abstract = "Knowledge-grounded dialogue generation aims to mitigate the issue of text degeneration by incorporating external knowledge to supplement the context. However, the model often fails to internalize this information into responses in a human-like manner. Instead, it simply inserts segments of the provided knowledge into generic responses. As a result, the generated responses tend to be tedious, incoherent, and in lack of interactivity which means the degeneration problem is still unsolved. In this work, we first find that such copying-style degeneration is primarily due to the weak likelihood objective, which allows the model to {``}cheat{''} the objective by merely duplicating knowledge segments in a superficial pattern matching based on overlap. To overcome this challenge, we then propose a Multi-level Adaptive Contrastive Learning (MACL) framework that dynamically samples negative examples and subsequently penalizes degeneration behaviors at both the token-level and sequence-level. Extensive experiments on the WoW dataset demonstrate the effectiveness of our approach across various pre-trained models and decoding strategies.", }

Knowledge-grounded dialogue generation aims to mitigate the issue of text degeneration by incorporating external knowledge to supplement the context. However, the model often fails to internalize this information into responses in a human-like manner. Instead, it simply inserts segments of the provided knowledge into generic responses. As a result, the generated responses tend to be tedious, incoherent, and in lack of interactivity which means the degeneration problem is still unsolved. In this work, we first find that such copying-style degeneration is primarily due to the weak likelihood objective, which allows the model to {``}cheat{''} the objective by merely duplicating knowledge segments in a superficial pattern matching based on overlap. To overcome this challenge, we then propose a Multi-level Adaptive Contrastive Learning (MACL) framework that dynamically samples negative examples and subsequently penalizes degeneration behaviors at both the token-level and sequence-level. Extensive experiments on the WoW dataset demonstrate the effectiveness of our approach across various pre-trained models and decoding strategies.

Multi-level Adaptive Contrastive Learning for Knowledge Internalization in Dialogue Generation

emnlp-main.497

2310.08943

https://huggingface.co/papers/2310.08943

Poster

https://aclanthology.org/2023.emnlp-main.498.bib

https://aclanthology.org/2023.emnlp-main.498/

@inproceedings{goldsack-etal-2023-enhancing, title = "Enhancing Biomedical Lay Summarisation with External Knowledge Graphs", author = "Goldsack, Tomas and Zhang, Zhihao and Tang, Chen and Scarton, Carolina and Lin, Chenghua", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.498", doi = "10.18653/v1/2023.emnlp-main.498", pages = "8016--8032", abstract = "Previous approaches for automatic lay summarisation are exclusively reliant on the source article that, given it is written for a technical audience (e.g., researchers), is unlikely to explicitly define all technical concepts or state all of the background information that is relevant for a lay audience. We address this issue by augmenting eLife, an existing biomedical lay summarisation dataset, with article-specific knowledge graphs, each containing detailed information on relevant biomedical concepts. Using both automatic and human evaluations, we systematically investigate the effectiveness of three different approaches for incorporating knowledge graphs within lay summarisation models, with each method targeting a distinct area of the encoder-decoder model architecture. Our results confirm that integrating graph-based domain knowledge can significantly benefit lay summarisation by substantially increasing the readability of generated text and improving the explanation of technical concepts.", }

Previous approaches for automatic lay summarisation are exclusively reliant on the source article that, given it is written for a technical audience (e.g., researchers), is unlikely to explicitly define all technical concepts or state all of the background information that is relevant for a lay audience. We address this issue by augmenting eLife, an existing biomedical lay summarisation dataset, with article-specific knowledge graphs, each containing detailed information on relevant biomedical concepts. Using both automatic and human evaluations, we systematically investigate the effectiveness of three different approaches for incorporating knowledge graphs within lay summarisation models, with each method targeting a distinct area of the encoder-decoder model architecture. Our results confirm that integrating graph-based domain knowledge can significantly benefit lay summarisation by substantially increasing the readability of generated text and improving the explanation of technical concepts.

Enhancing Biomedical Lay Summarisation with External Knowledge Graphs

emnlp-main.498

2310.15702

Poster

https://aclanthology.org/2023.emnlp-main.499.bib

https://aclanthology.org/2023.emnlp-main.499/

@inproceedings{shi-etal-2023-diffusion, title = "A Diffusion Weighted Graph Framework for New Intent Discovery", author = "Shi, Wenkai and An, Wenbin and Tian, Feng and Zheng, Qinghua and Wang, QianYing and Chen, Ping", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.499", doi = "10.18653/v1/2023.emnlp-main.499", pages = "8033--8042", abstract = "New Intent Discovery (NID) aims to recognize both new and known intents from unlabeled data with the aid of limited labeled data containing only known intents. Without considering structure relationships between samples, previous methods generate noisy supervisory signals which cannot strike a balance between quantity and quality, hindering the formation of new intent clusters and effective transfer of the pre-training knowledge. To mitigate this limitation, we propose a novel $\textit{Diffusion Weighted Graph Framework}$ (DWGF) to capture both semantic similarities and structure relationships inherent in data, enabling more sufficient and reliable supervisory signals. Specifically, for each sample, we diffuse neighborhood relationships along semantic paths guided by the nearest neighbors for multiple hops to characterize its local structure discriminately. Then, we sample its positive keys and weigh them based on semantic similarities and local structures for contrastive learning. During inference, we further propose $\textit{Graph Smoothing Filter}$ (GSF) to explicitly utilize the structure relationships to filter high-frequency noise embodied in semantically ambiguous samples on the cluster boundary. Extensive experiments show that our method outperforms state-of-the-art models on all evaluation metrics across multiple benchmark datasets. Code and data will be made public.", }

New Intent Discovery (NID) aims to recognize both new and known intents from unlabeled data with the aid of limited labeled data containing only known intents. Without considering structure relationships between samples, previous methods generate noisy supervisory signals which cannot strike a balance between quantity and quality, hindering the formation of new intent clusters and effective transfer of the pre-training knowledge. To mitigate this limitation, we propose a novel $\textit{Diffusion Weighted Graph Framework}$ (DWGF) to capture both semantic similarities and structure relationships inherent in data, enabling more sufficient and reliable supervisory signals. Specifically, for each sample, we diffuse neighborhood relationships along semantic paths guided by the nearest neighbors for multiple hops to characterize its local structure discriminately. Then, we sample its positive keys and weigh them based on semantic similarities and local structures for contrastive learning. During inference, we further propose $\textit{Graph Smoothing Filter}$ (GSF) to explicitly utilize the structure relationships to filter high-frequency noise embodied in semantically ambiguous samples on the cluster boundary. Extensive experiments show that our method outperforms state-of-the-art models on all evaluation metrics across multiple benchmark datasets. Code and data will be made public.

A Diffusion Weighted Graph Framework for New Intent Discovery

emnlp-main.499

2310.15836

Poster

https://aclanthology.org/2023.emnlp-main.500.bib

https://aclanthology.org/2023.emnlp-main.500/

@inproceedings{nguyen-etal-2023-self, title = "A Self-enhancement Multitask Framework for Unsupervised Aspect Category Detection", author = "Nguyen, Thi-Nhung and Ngo, Hoang and Nguyen, Kiem-Hieu and Cao, Tuan-Dung", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.500", doi = "10.18653/v1/2023.emnlp-main.500", pages = "8043--8054", abstract = "Our work addresses the problem of unsupervised Aspect Category Detection using a small set of seed words. Recent works have focused on learning embedding spaces for seed words and sentences to establish similarities between sentences and aspects. However, aspect representations are limited by the quality of initial seed words, and model performances are compromised by noise. To mitigate this limitation, we propose a simple framework that automatically enhances the quality of initial seed words and selects high-quality sentences for training instead of using the entire dataset. Our main concepts are to add a number of seed words to the initial set and to treat the task of noise resolution as a task of augmenting data for a low-resource task. In addition, we jointly train Aspect Category Detection with Aspect Term Extraction and Aspect Term Polarity to further enhance performance. This approach facilitates shared representation learning, allowing Aspect Category Detection to benefit from the additional guidance offered by other tasks. Extensive experiments demonstrate that our framework surpasses strong baselines on standard datasets.", }

Our work addresses the problem of unsupervised Aspect Category Detection using a small set of seed words. Recent works have focused on learning embedding spaces for seed words and sentences to establish similarities between sentences and aspects. However, aspect representations are limited by the quality of initial seed words, and model performances are compromised by noise. To mitigate this limitation, we propose a simple framework that automatically enhances the quality of initial seed words and selects high-quality sentences for training instead of using the entire dataset. Our main concepts are to add a number of seed words to the initial set and to treat the task of noise resolution as a task of augmenting data for a low-resource task. In addition, we jointly train Aspect Category Detection with Aspect Term Extraction and Aspect Term Polarity to further enhance performance. This approach facilitates shared representation learning, allowing Aspect Category Detection to benefit from the additional guidance offered by other tasks. Extensive experiments demonstrate that our framework surpasses strong baselines on standard datasets.

A Self-enhancement Multitask Framework for Unsupervised Aspect Category Detection

emnlp-main.500

2311.09708

Poster