Datasets:

EMNLP

/

EMNLP2023-papers

like 0

Follow

Conference on Empirical Methods in Natural Language Processing 1

https://aclanthology.org/2023.findings-emnlp.820.bib

https://aclanthology.org/2023.findings-emnlp.820/

@inproceedings{zhang-etal-2023-pre, title = "Pre-training Multi-task Contrastive Learning Models for Scientific Literature Understanding", author = "Zhang, Yu and Cheng, Hao and Shen, Zhihong and Liu, Xiaodong and Wang, Ye-Yi and Gao, Jianfeng", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.820", doi = "10.18653/v1/2023.findings-emnlp.820", pages = "12259--12275", abstract = "Scientific literature understanding tasks have gained significant attention due to their potential to accelerate scientific discovery. Pre-trained language models (LMs) have shown effectiveness in these tasks, especially when tuned via contrastive learning. However, jointly utilizing pre-training data across multiple heterogeneous tasks (e.g., extreme multi-label paper classification, citation prediction, and literature search) remains largely unexplored. To bridge this gap, we propose a multi-task contrastive learning framework, SciMult, with a focus on facilitating common knowledge sharing across different scientific literature understanding tasks while preventing task-specific skills from interfering with each other. To be specific, we explore two techniques {--} task-aware specialization and instruction tuning. The former adopts a Mixture-of-Experts Transformer architecture with task-aware sub-layers; the latter prepends task-specific instructions to the input text so as to produce task-aware outputs. Extensive experiments on a comprehensive collection of benchmark datasets verify the effectiveness of our task-aware specialization strategy, where we outperform state-of-the-art scientific pre-trained LMs. Code, datasets, and pre-trained models can be found at https://scimult.github.io/.", }

Scientific literature understanding tasks have gained significant attention due to their potential to accelerate scientific discovery. Pre-trained language models (LMs) have shown effectiveness in these tasks, especially when tuned via contrastive learning. However, jointly utilizing pre-training data across multiple heterogeneous tasks (e.g., extreme multi-label paper classification, citation prediction, and literature search) remains largely unexplored. To bridge this gap, we propose a multi-task contrastive learning framework, SciMult, with a focus on facilitating common knowledge sharing across different scientific literature understanding tasks while preventing task-specific skills from interfering with each other. To be specific, we explore two techniques {--} task-aware specialization and instruction tuning. The former adopts a Mixture-of-Experts Transformer architecture with task-aware sub-layers; the latter prepends task-specific instructions to the input text so as to produce task-aware outputs. Extensive experiments on a comprehensive collection of benchmark datasets verify the effectiveness of our task-aware specialization strategy, where we outperform state-of-the-art scientific pre-trained LMs. Code, datasets, and pre-trained models can be found at https://scimult.github.io/.

Pre-training Multi-task Contrastive Learning Models for Scientific Literature Understanding

findings-emnlp.820

2305.14232

Poster

https://aclanthology.org/2023.findings-emnlp.821.bib

https://aclanthology.org/2023.findings-emnlp.821/

@inproceedings{samo-etal-2023-blm, title = "{BLM}-s/l{E}: A structured dataset of {E}nglish spray-load verb alternations for testing generalization in {LLM}s", author = "Samo, Giuseppe and Nastase, Vivi and Jiang, Chunyang and Merlo, Paola", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.821", doi = "10.18653/v1/2023.findings-emnlp.821", pages = "12276--12287", abstract = "Current NLP models appear to be achieving performance comparable to human capabilities on well-established benchmarks. New benchmarks are now necessary to test deeper layers of understanding of natural languages by these models. Blackbird{'}s Language Matrices are a recently developed framework that draws inspiration from tests of human analytic intelligence. The BLM task has revealed that successful performances in previously studied linguistic problems do not yet stem from a deep understanding of the generative factors that define these problems. In this study, we define a new BLM task for predicate-argument structure, and develop a structured dataset for its investigation, concentrating on the spray-load verb alternations in English, as a case study. The context sentences include one alternant from the spray-load alternation and the target sentence is the other alternant, to be chosen among a minimally contrastive and adversarial set of answers. We describe the generation process of the dataset and the reasoning behind the generating rules. The dataset aims to facilitate investigations into how verb information is encoded in sentence embeddings and how models generalize to the complex properties of argument structures. Benchmarking experiments conducted on the dataset and qualitative error analysis on the answer set reveal the inherent challenges associated with the problem even for current high-performing representations.", }

Current NLP models appear to be achieving performance comparable to human capabilities on well-established benchmarks. New benchmarks are now necessary to test deeper layers of understanding of natural languages by these models. Blackbird{'}s Language Matrices are a recently developed framework that draws inspiration from tests of human analytic intelligence. The BLM task has revealed that successful performances in previously studied linguistic problems do not yet stem from a deep understanding of the generative factors that define these problems. In this study, we define a new BLM task for predicate-argument structure, and develop a structured dataset for its investigation, concentrating on the spray-load verb alternations in English, as a case study. The context sentences include one alternant from the spray-load alternation and the target sentence is the other alternant, to be chosen among a minimally contrastive and adversarial set of answers. We describe the generation process of the dataset and the reasoning behind the generating rules. The dataset aims to facilitate investigations into how verb information is encoded in sentence embeddings and how models generalize to the complex properties of argument structures. Benchmarking experiments conducted on the dataset and qualitative error analysis on the answer set reveal the inherent challenges associated with the problem even for current high-performing representations.

BLM-s/lE: A structured dataset of English spray-load verb alternations for testing generalization in LLMs

findings-emnlp.821

Poster

https://aclanthology.org/2023.findings-emnlp.822.bib

https://aclanthology.org/2023.findings-emnlp.822/

@inproceedings{ye-etal-2023-efficiently, title = "Efficiently Enhancing Zero-Shot Performance of Instruction Following Model via Retrieval of Soft Prompt", author = "Ye, Seonghyeon and Jang, Joel and Kim, Doyoung and Jo, Yongrae and Seo, Minjoon", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.822", doi = "10.18653/v1/2023.findings-emnlp.822", pages = "12288--12309", abstract = "Enhancing the zero-shot performance of instruction-following models requires heavy computation, either by scaling the total number of training datasets or the model size. In this work, we explore how retrieval of soft prompts obtained through prompt tuning can efficiently assist hard prompts in zero-shot task generalization. Specifically, we train soft prompt embeddings for each prompt through prompt tuning, store the samples of the training instances mapped with the prompt embeddings, and retrieve the corresponding prompt embedding of the training instance closest to the query instance during inference. While only adding 0.007{\%} additional parameters, retrieval of soft prompt enhances the performance of T0 on unseen tasks by outperforming it on 10 out of 11 datasets as well as improving the mean accuracy of T0 on BIG-bench benchmark by 2.39{\%} points. Also, we report an interesting finding that retrieving source embeddings trained on similar answer choice formats is more important than those on similar task types.", }

Enhancing the zero-shot performance of instruction-following models requires heavy computation, either by scaling the total number of training datasets or the model size. In this work, we explore how retrieval of soft prompts obtained through prompt tuning can efficiently assist hard prompts in zero-shot task generalization. Specifically, we train soft prompt embeddings for each prompt through prompt tuning, store the samples of the training instances mapped with the prompt embeddings, and retrieve the corresponding prompt embedding of the training instance closest to the query instance during inference. While only adding 0.007{\%} additional parameters, retrieval of soft prompt enhances the performance of T0 on unseen tasks by outperforming it on 10 out of 11 datasets as well as improving the mean accuracy of T0 on BIG-bench benchmark by 2.39{\%} points. Also, we report an interesting finding that retrieving source embeddings trained on similar answer choice formats is more important than those on similar task types.

Efficiently Enhancing Zero-Shot Performance of Instruction Following Model via Retrieval of Soft Prompt

findings-emnlp.822

2210.03029

Poster

https://aclanthology.org/2023.findings-emnlp.823.bib

https://aclanthology.org/2023.findings-emnlp.823/

@inproceedings{schwobel-etal-2023-geographical, title = "Geographical Erasure in Language Generation", author = {Schw{\"o}bel, Pola and Golebiowski, Jacek and Donini, Michele and Archambeau, Cedric and Pruthi, Danish}, editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.823", doi = "10.18653/v1/2023.findings-emnlp.823", pages = "12310--12324", abstract = "Large language models (LLMs) encode vast amounts of world knowledge. However, since these models are trained on large swaths of internet data, they are at risk of inordinately capturing information about dominant groups. This imbalance can propagate into generated language. In this work, we study and operationalise a form of geographical erasure wherein language models underpredict certain countries. We demonstrate consistent instances of erasure across a range of LLMs. We discover that erasure strongly correlates with low frequencies of country mentions in the training corpus. Lastly, we mitigate erasure by finetuning using a custom objective.", }

Large language models (LLMs) encode vast amounts of world knowledge. However, since these models are trained on large swaths of internet data, they are at risk of inordinately capturing information about dominant groups. This imbalance can propagate into generated language. In this work, we study and operationalise a form of geographical erasure wherein language models underpredict certain countries. We demonstrate consistent instances of erasure across a range of LLMs. We discover that erasure strongly correlates with low frequencies of country mentions in the training corpus. Lastly, we mitigate erasure by finetuning using a custom objective.

Geographical Erasure in Language Generation

findings-emnlp.823

2310.14777

Poster

https://aclanthology.org/2023.findings-emnlp.824.bib

https://aclanthology.org/2023.findings-emnlp.824/

@inproceedings{bao-etal-2023-foundation, title = "Can Foundation Models Watch, Talk and Guide You Step by Step to Make a Cake?", author = "Bao, Yuwei and Yu, Keunwoo and Zhang, Yichi and Storks, Shane and Bar-Yossef, Itamar and de la Iglesia, Alex and Su, Megan and Zheng, Xiao and Chai, Joyce", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.824", doi = "10.18653/v1/2023.findings-emnlp.824", pages = "12325--12341", abstract = "Despite tremendous advances in AI, it remains a significant challenge to develop interactive task guidance systems that can offer situated, personalized guidance and assist humans in various tasks. These systems need to have a sophisticated understanding of the user as well as the environment, and make timely accurate decisions on when and what to say. To address this issue, we created a new multimodal benchmark dataset, Watch, Talk and Guide (WTaG) based on natural interaction between a human user and a human instructor. We further proposed two tasks: User and Environment Understanding, and Instructor Decision Making. We leveraged several foundation models to study to what extent these models can be quickly adapted to perceptually enabled task guidance. Our quantitative, qualitative, and human evaluation results show that these models can demonstrate fair performances in some cases with no task-specific training, but a fast and reliable adaptation remains a significant challenge. Our benchmark and baselines will provide a stepping stone for future work on situated task guidance.", }

Despite tremendous advances in AI, it remains a significant challenge to develop interactive task guidance systems that can offer situated, personalized guidance and assist humans in various tasks. These systems need to have a sophisticated understanding of the user as well as the environment, and make timely accurate decisions on when and what to say. To address this issue, we created a new multimodal benchmark dataset, Watch, Talk and Guide (WTaG) based on natural interaction between a human user and a human instructor. We further proposed two tasks: User and Environment Understanding, and Instructor Decision Making. We leveraged several foundation models to study to what extent these models can be quickly adapted to perceptually enabled task guidance. Our quantitative, qualitative, and human evaluation results show that these models can demonstrate fair performances in some cases with no task-specific training, but a fast and reliable adaptation remains a significant challenge. Our benchmark and baselines will provide a stepping stone for future work on situated task guidance.

Can Foundation Models Watch, Talk and Guide You Step by Step to Make a Cake?

findings-emnlp.824

2311.00738

Poster

https://aclanthology.org/2023.findings-emnlp.825.bib

https://aclanthology.org/2023.findings-emnlp.825/

@inproceedings{tay-etal-2023-scaling, title = "Scaling Laws vs Model Architectures: How does Inductive Bias Influence Scaling?", author = "Tay, Yi and Dehghani, Mostafa and Abnar, Samira and Chung, Hyung and Fedus, William and Rao, Jinfeng and Narang, Sharan and Tran, Vinh and Yogatama, Dani and Metzler, Donald", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.825", doi = "10.18653/v1/2023.findings-emnlp.825", pages = "12342--12364", abstract = "There have been a lot of interest in the scaling properties of Transformer models. However, not much has been done on the front of investigating the effect of scaling properties of different inductive biases and model architectures. Do model architectures scale differently? If so, how does inductive bias affect scaling behaviour? How does this influence upstream (pretraining) and downstream (transfer)? This paper conducts a systematic study of scaling behaviour of ten diverse model architectures such as Transformers, Switch Transformers, Universal Transformers, Dynamic convolutions, Performers, and recently proposed MLP-Mixers. Via extensive experiments, we show that (1) architecture is an indeed an important consideration when performing scaling and (2) the best performing model can fluctuate at different scales. We believe that the findings outlined in this work has significant implications to how model architectures are currently evaluated in the community.", }

There have been a lot of interest in the scaling properties of Transformer models. However, not much has been done on the front of investigating the effect of scaling properties of different inductive biases and model architectures. Do model architectures scale differently? If so, how does inductive bias affect scaling behaviour? How does this influence upstream (pretraining) and downstream (transfer)? This paper conducts a systematic study of scaling behaviour of ten diverse model architectures such as Transformers, Switch Transformers, Universal Transformers, Dynamic convolutions, Performers, and recently proposed MLP-Mixers. Via extensive experiments, we show that (1) architecture is an indeed an important consideration when performing scaling and (2) the best performing model can fluctuate at different scales. We believe that the findings outlined in this work has significant implications to how model architectures are currently evaluated in the community.

Scaling Laws vs Model Architectures: How does Inductive Bias Influence Scaling?

findings-emnlp.825

https://huggingface.co/papers/2207.10551

Poster

https://aclanthology.org/2023.findings-emnlp.826.bib

https://aclanthology.org/2023.findings-emnlp.826/

@inproceedings{huang-etal-2023-languages, title = "Not All Languages Are Created Equal in {LLM}s: Improving Multilingual Capability by Cross-Lingual-Thought Prompting", author = "Huang, Haoyang and Tang, Tianyi and Zhang, Dongdong and Zhao, Xin and Song, Ting and Xia, Yan and Wei, Furu", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.826", doi = "10.18653/v1/2023.findings-emnlp.826", pages = "12365--12394", abstract = "Large language models (LLMs) demonstrate impressive multilingual capability, but their performance varies substantially across different languages. In this work, we introduce a simple yet effective method, called cross-lingual-thought prompting (XLT), to systematically improve the multilingual capability of LLMs. Specifically, XLT is a generic template prompt that stimulates cross-lingual and logical reasoning skills to enhance task performance across languages. We conduct comprehensive evaluations on 7 typical benchmarks related to reasoning, understanding, and generation tasks, covering both high-resource and low-resource languages. Experimental results show that XLT not only remarkably enhances the performance of various multilingual tasks but also significantly reduces the gap between the average performance and the best performance of each task in different languages. Notably, XLT brings over 10 points of average improvement in arithmetic reasoning and open-domain question-answering tasks.", }

Large language models (LLMs) demonstrate impressive multilingual capability, but their performance varies substantially across different languages. In this work, we introduce a simple yet effective method, called cross-lingual-thought prompting (XLT), to systematically improve the multilingual capability of LLMs. Specifically, XLT is a generic template prompt that stimulates cross-lingual and logical reasoning skills to enhance task performance across languages. We conduct comprehensive evaluations on 7 typical benchmarks related to reasoning, understanding, and generation tasks, covering both high-resource and low-resource languages. Experimental results show that XLT not only remarkably enhances the performance of various multilingual tasks but also significantly reduces the gap between the average performance and the best performance of each task in different languages. Notably, XLT brings over 10 points of average improvement in arithmetic reasoning and open-domain question-answering tasks.

Not All Languages Are Created Equal in LLMs: Improving Multilingual Capability by Cross-Lingual-Thought Prompting

findings-emnlp.826

2305.07004

https://huggingface.co/papers/2305.07004

Poster

https://aclanthology.org/2023.findings-emnlp.827.bib

https://aclanthology.org/2023.findings-emnlp.827/

@inproceedings{su-etal-2023-detectllm, title = "{D}etect{LLM}: Leveraging Log Rank Information for Zero-Shot Detection of Machine-Generated Text", author = "Su, Jinyan and Zhuo, Terry and Wang, Di and Nakov, Preslav", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.827", doi = "10.18653/v1/2023.findings-emnlp.827", pages = "12395--12412", abstract = "With the rapid progress of Large language models (LLMs) and the huge amount of text they generate, it becomes impractical to manually distinguish whether a text is machine-generated. The growing use of LLMs in social media and education, prompts us to develop methods to detect machine-generated text, preventing malicious use such as plagiarism, misinformation, and propaganda. In this paper, we introduce two novel zero-shot methods for detecting machine-generated text by leveraging the Log-Rank information. One is called DetectLLM-LRR, which is fast and efficient, and the other is called DetectLLM-NPR, which is more accurate, but slower due to the need for perturbations. Our experiments on three datasets and seven language models show that our proposed methods improve over the state of the art by 3.9 and 1.75 AUROC points absolute. Moreover, DetectLLM-NPR needs fewer perturbations than previous work to achieve the same level of performance, which makes it more practical for real-world use. We also investigate the efficiency-performance trade-off based on users{'} preference for these two measures and provide intuition for using them in practice effectively. We release the data and the code of both methods in https://github.com/mbzuai-nlp/DetectLLM.", }

With the rapid progress of Large language models (LLMs) and the huge amount of text they generate, it becomes impractical to manually distinguish whether a text is machine-generated. The growing use of LLMs in social media and education, prompts us to develop methods to detect machine-generated text, preventing malicious use such as plagiarism, misinformation, and propaganda. In this paper, we introduce two novel zero-shot methods for detecting machine-generated text by leveraging the Log-Rank information. One is called DetectLLM-LRR, which is fast and efficient, and the other is called DetectLLM-NPR, which is more accurate, but slower due to the need for perturbations. Our experiments on three datasets and seven language models show that our proposed methods improve over the state of the art by 3.9 and 1.75 AUROC points absolute. Moreover, DetectLLM-NPR needs fewer perturbations than previous work to achieve the same level of performance, which makes it more practical for real-world use. We also investigate the efficiency-performance trade-off based on users{'} preference for these two measures and provide intuition for using them in practice effectively. We release the data and the code of both methods in https://github.com/mbzuai-nlp/DetectLLM.

DetectLLM: Leveraging Log Rank Information for Zero-Shot Detection of Machine-Generated Text

findings-emnlp.827

2306.05540

https://huggingface.co/papers/2306.05540

Poster

https://aclanthology.org/2023.findings-emnlp.828.bib

https://aclanthology.org/2023.findings-emnlp.828/

@inproceedings{junbing-etal-2023-complex, title = "From Complex to Simple: Unraveling the Cognitive Tree for Reasoning with Small Language Models", author = "Junbing, Yan and Wang, Chengyu and Zhang, Taolin and He, Xiaofeng and Huang, Jun and Zhang, Wei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.828", doi = "10.18653/v1/2023.findings-emnlp.828", pages = "12413--12425", abstract = "Reasoning is a distinctive human capacity, enabling us to address complex problems by breaking them down into a series of manageable cognitive steps. Yet, complex logical reasoning is still cumbersome for language models. Based on the dual process theory in cognitive science, we are the first to unravel the cognitive reasoning abilities of language models. Our framework employs an iterative methodology to construct a Cognitive Tree (CogTree). The root node of this tree represents the initial query, while the leaf nodes consist of straightforward questions that can be answered directly. This construction involves two main components: the implicit extraction module (referred to as the intuitive system) and the explicit reasoning module (referred to as the reflective system). The intuitive system rapidly generates multiple responses by utilizing in-context examples, while the reflective system scores these responses using comparative learning. The scores guide the intuitive system in its subsequent generation step.Our experimental results on two popular and challenging reasoning tasks indicate that it is possible to achieve a performance level comparable to that of GPT-3.5 (with 175B parameters), using a significantly smaller language model that contains fewer parameters ({\textless}=7B) than 5{\%} of GPT-3.5.", }

Reasoning is a distinctive human capacity, enabling us to address complex problems by breaking them down into a series of manageable cognitive steps. Yet, complex logical reasoning is still cumbersome for language models. Based on the dual process theory in cognitive science, we are the first to unravel the cognitive reasoning abilities of language models. Our framework employs an iterative methodology to construct a Cognitive Tree (CogTree). The root node of this tree represents the initial query, while the leaf nodes consist of straightforward questions that can be answered directly. This construction involves two main components: the implicit extraction module (referred to as the intuitive system) and the explicit reasoning module (referred to as the reflective system). The intuitive system rapidly generates multiple responses by utilizing in-context examples, while the reflective system scores these responses using comparative learning. The scores guide the intuitive system in its subsequent generation step.Our experimental results on two popular and challenging reasoning tasks indicate that it is possible to achieve a performance level comparable to that of GPT-3.5 (with 175B parameters), using a significantly smaller language model that contains fewer parameters ({\textless}=7B) than 5{\%} of GPT-3.5.

From Complex to Simple: Unraveling the Cognitive Tree for Reasoning with Small Language Models

findings-emnlp.828

2311.06754

https://huggingface.co/papers/2311.06754

Poster

https://aclanthology.org/2023.findings-emnlp.829.bib

https://aclanthology.org/2023.findings-emnlp.829/

@inproceedings{wang-etal-2023-macedon, title = "Macedon: Minimizing Representation Coding Rate Reduction for Cross-Lingual Natural Language Understanding", author = "Wang, Haoyu and Wang, Yaqing and Yao, Huaxiu and Gao, Jing", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.829", doi = "10.18653/v1/2023.findings-emnlp.829", pages = "12426--12436", abstract = "Cross-lingual natural language understanding(NLU) is one of the fundamental tasks of NLP. The goal is to learn a model which can generalize well on both high-resource and low-resource language data. Recent pre-trained multilingual language models, e.g., multilingual BERT, XLM, have shown impressive performance on cross-lingual NLU tasks. However, such promising results request the use of sufficient training data, which is a difficult condition to satisfy for low-resource language. When the data is limited in those low resource languages, the accuracy of existing models will drop. In light of this challenge, we investigate the important task of how to train the cross-lingual model with abundant high-source language data and limited low-resource language data. Existing methods typically learn language-agnostic representation via adversarial training and mutual information estimation. Existing approaches may suffer When data is very limited (e.g., low-resource language) because it is challenging to estimate data distribution accurately. To tackle this issue, we propose a conceptually innovative approach to remove language-associated information via \textbf{m}inimizing represent\textbf{a}tion \textbf{c}oding rate r\textbf{ed}ucti\textbf{on}(Macedon). Specifically, Macedon avoids using extra codes to encode language-related information, which is measured by the rate-distortion function. To validate the effectiveness of Macedon, we conduct extensive experiments on three tasks, including paraphrase identification, natural language inference, and query advertisement matching. The experiment results show that the proposed Macedon outperforms state-of-the-art cross-lingual NLU approaches.", }

Cross-lingual natural language understanding(NLU) is one of the fundamental tasks of NLP. The goal is to learn a model which can generalize well on both high-resource and low-resource language data. Recent pre-trained multilingual language models, e.g., multilingual BERT, XLM, have shown impressive performance on cross-lingual NLU tasks. However, such promising results request the use of sufficient training data, which is a difficult condition to satisfy for low-resource language. When the data is limited in those low resource languages, the accuracy of existing models will drop. In light of this challenge, we investigate the important task of how to train the cross-lingual model with abundant high-source language data and limited low-resource language data. Existing methods typically learn language-agnostic representation via adversarial training and mutual information estimation. Existing approaches may suffer When data is very limited (e.g., low-resource language) because it is challenging to estimate data distribution accurately. To tackle this issue, we propose a conceptually innovative approach to remove language-associated information via \textbf{m}inimizing represent\textbf{a}tion \textbf{c}oding rate r\textbf{ed}ucti\textbf{on}(Macedon). Specifically, Macedon avoids using extra codes to encode language-related information, which is measured by the rate-distortion function. To validate the effectiveness of Macedon, we conduct extensive experiments on three tasks, including paraphrase identification, natural language inference, and query advertisement matching. The experiment results show that the proposed Macedon outperforms state-of-the-art cross-lingual NLU approaches.

Macedon: Minimizing Representation Coding Rate Reduction for Cross-Lingual Natural Language Understanding

findings-emnlp.829

Poster

https://aclanthology.org/2023.findings-emnlp.830.bib

https://aclanthology.org/2023.findings-emnlp.830/

@inproceedings{jin-etal-2023-adversarial, title = "Adversarial Robustness for Large Language {NER} models using Disentanglement and Word Attributions", author = "Jin, Xiaomeng and Vinzamuri, Bhanukiran and Venkatapathy, Sriram and Ji, Heng and Natarajan, Pradeep", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.830", doi = "10.18653/v1/2023.findings-emnlp.830", pages = "12437--12450", abstract = "Large language models (LLM{'}s) have been widely used for several applications such as question answering, text classification and clustering. While the preliminary results across the aforementioned tasks looks promising, recent work has dived deep into LLM{'}s performing poorly for complex Named Entity Recognition (NER) tasks in comparison to fine-tuned pre-trained language models (PLM{'}s). To enhance wider adoption of LLM{'}s, our paper investigates the robustness of such LLM NER models and its instruction fine-tuned variants to adversarial attacks. In particular, we propose a novel attack which relies on disentanglement and word attribution techniques where the former aids in learning an embedding capturing both entity and non-entity influences separately, and the latter aids in identifying important words across both components. This is in stark contrast to most techniques which primarily leverage non-entity words for perturbations limiting the space being explored to synthesize effective adversarial examples. Adversarial training results based on our method improves the F1 score over original LLM NER model by 8{\%} and 18{\%} on CoNLL-2003 and Ontonotes 5.0 datasets respectively.", }

Large language models (LLM{'}s) have been widely used for several applications such as question answering, text classification and clustering. While the preliminary results across the aforementioned tasks looks promising, recent work has dived deep into LLM{'}s performing poorly for complex Named Entity Recognition (NER) tasks in comparison to fine-tuned pre-trained language models (PLM{'}s). To enhance wider adoption of LLM{'}s, our paper investigates the robustness of such LLM NER models and its instruction fine-tuned variants to adversarial attacks. In particular, we propose a novel attack which relies on disentanglement and word attribution techniques where the former aids in learning an embedding capturing both entity and non-entity influences separately, and the latter aids in identifying important words across both components. This is in stark contrast to most techniques which primarily leverage non-entity words for perturbations limiting the space being explored to synthesize effective adversarial examples. Adversarial training results based on our method improves the F1 score over original LLM NER model by 8{\%} and 18{\%} on CoNLL-2003 and Ontonotes 5.0 datasets respectively.

Adversarial Robustness for Large Language NER models using Disentanglement and Word Attributions

findings-emnlp.830

Poster

https://aclanthology.org/2023.findings-emnlp.831.bib

https://aclanthology.org/2023.findings-emnlp.831/

@inproceedings{vats-etal-2023-llms, title = "{LLM}s {--} the Good, the Bad or the Indispensable?: A Use Case on Legal Statute Prediction and Legal Judgment Prediction on {I}ndian Court Cases", author = "Vats, Shaurya and Zope, Atharva and De, Somsubhra and Sharma, Anurag and Bhattacharya, Upal and Nigam, Shubham and Guha, Shouvik and Rudra, Koustav and Ghosh, Kripabandhu", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.831", doi = "10.18653/v1/2023.findings-emnlp.831", pages = "12451--12474", abstract = "The Large Language Models (LLMs) have impacted many real-life tasks. To examine the efficacy of LLMs in a high-stake domain like law, we have applied state-of-the-art LLMs for two popular tasks: Statute Prediction and Judgment Prediction, on Indian Supreme Court cases. We see that while LLMs exhibit excellent predictive performance in Statute Prediction, their performance dips in Judgment Prediction when compared with many standard models. The explanations generated by LLMs (along with prediction) are of moderate to decent quality. We also see evidence of gender and religious bias in the LLM-predicted results. In addition, we present a note from a senior legal expert on the ethical concerns of deploying LLMs in these critical legal tasks.", }

The Large Language Models (LLMs) have impacted many real-life tasks. To examine the efficacy of LLMs in a high-stake domain like law, we have applied state-of-the-art LLMs for two popular tasks: Statute Prediction and Judgment Prediction, on Indian Supreme Court cases. We see that while LLMs exhibit excellent predictive performance in Statute Prediction, their performance dips in Judgment Prediction when compared with many standard models. The explanations generated by LLMs (along with prediction) are of moderate to decent quality. We also see evidence of gender and religious bias in the LLM-predicted results. In addition, we present a note from a senior legal expert on the ethical concerns of deploying LLMs in these critical legal tasks.

LLMs – the Good, the Bad or the Indispensable?: A Use Case on Legal Statute Prediction and Legal Judgment Prediction on Indian Court Cases

findings-emnlp.831

Poster

https://aclanthology.org/2023.findings-emnlp.832.bib

https://aclanthology.org/2023.findings-emnlp.832/

@inproceedings{deng-etal-2023-annotate, title = "You Are What You Annotate: Towards Better Models through Annotator Representations", author = "Deng, Naihao and Zhang, Xinliang and Liu, Siyang and Wu, Winston and Wang, Lu and Mihalcea, Rada", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.832", doi = "10.18653/v1/2023.findings-emnlp.832", pages = "12475--12498", abstract = "Annotator disagreement is ubiquitous in natural language processing (NLP) tasks. There are multiple reasons for such disagreements, including the subjectivity of the task, difficult cases, unclear guidelines, and so on. Rather than simply aggregating labels to obtain data annotations, we instead try to directly model the diverse perspectives of the annotators, and explicitly account for annotators{'} idiosyncrasies in the modeling process by creating representations for each annotator (*annotator embeddings*) and also their annotations (*annotation embeddings*). In addition, we propose **TID-8**, **T**he **I**nherent **D**isagreement - **8** dataset, a benchmark that consists of eight existing language understanding datasets that have inherent annotator disagreement. We test our approach on TID-8 and show that our approach helps models learn significantly better from disagreements on six different datasets in TID-8 while increasing model size by fewer than 1{\%} parameters. By capturing the unique tendencies and subjectivity of individual annotators through embeddings, our representations prime AI models to be inclusive of diverse viewpoints.", }

Annotator disagreement is ubiquitous in natural language processing (NLP) tasks. There are multiple reasons for such disagreements, including the subjectivity of the task, difficult cases, unclear guidelines, and so on. Rather than simply aggregating labels to obtain data annotations, we instead try to directly model the diverse perspectives of the annotators, and explicitly account for annotators{'} idiosyncrasies in the modeling process by creating representations for each annotator (*annotator embeddings*) and also their annotations (*annotation embeddings*). In addition, we propose **TID-8**, **T**he **I**nherent **D**isagreement - **8** dataset, a benchmark that consists of eight existing language understanding datasets that have inherent annotator disagreement. We test our approach on TID-8 and show that our approach helps models learn significantly better from disagreements on six different datasets in TID-8 while increasing model size by fewer than 1{\%} parameters. By capturing the unique tendencies and subjectivity of individual annotators through embeddings, our representations prime AI models to be inclusive of diverse viewpoints.

You Are What You Annotate: Towards Better Models through Annotator Representations

findings-emnlp.832

2305.14663

https://huggingface.co/papers/2305.14663

Poster

https://aclanthology.org/2023.findings-emnlp.833.bib

https://aclanthology.org/2023.findings-emnlp.833/

@inproceedings{you-etal-2023-large, title = "Large Language Models Are Better Adversaries: Exploring Generative Clean-Label Backdoor Attacks Against Text Classifiers", author = "You, Wencong and Hammoudeh, Zayd and Lowd, Daniel", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.833", doi = "10.18653/v1/2023.findings-emnlp.833", pages = "12499--12527", abstract = "Backdoor attacks manipulate model predictions by inserting innocuous triggers into training and test data. We focus on more realistic and more challenging clean-label attacks where the adversarial training examples are correctly labeled. Our attack, LLMBkd, leverages language models to automatically insert diverse style-based triggers into texts. We also propose a poison selection technique to improve the effectiveness of both LLMBkd as well as existing textual backdoor attacks. Lastly, we describe REACT, a baseline defense to mitigate backdoor attacks via antidote training examples. Our evaluations demonstrate LLMBkd{'}s effectiveness and efficiency, where we consistently achieve high attack success rates across a wide range of styles with little effort and no model training.", }

Backdoor attacks manipulate model predictions by inserting innocuous triggers into training and test data. We focus on more realistic and more challenging clean-label attacks where the adversarial training examples are correctly labeled. Our attack, LLMBkd, leverages language models to automatically insert diverse style-based triggers into texts. We also propose a poison selection technique to improve the effectiveness of both LLMBkd as well as existing textual backdoor attacks. Lastly, we describe REACT, a baseline defense to mitigate backdoor attacks via antidote training examples. Our evaluations demonstrate LLMBkd{'}s effectiveness and efficiency, where we consistently achieve high attack success rates across a wide range of styles with little effort and no model training.

Large Language Models Are Better Adversaries: Exploring Generative Clean-Label Backdoor Attacks Against Text Classifiers

findings-emnlp.833

2310.18603

Poster

https://aclanthology.org/2023.findings-emnlp.834.bib

https://aclanthology.org/2023.findings-emnlp.834/

@inproceedings{wang-etal-2023-noise, title = "Noise-Robust Fine-Tuning of Pretrained Language Models via External Guidance", author = "Wang, Song and Tan, Zhen and Guo, Ruocheng and Li, Jundong", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.834", doi = "10.18653/v1/2023.findings-emnlp.834", pages = "12528--12540", abstract = "Adopting a two-stage paradigm of pretraining followed by fine-tuning, Pretrained Language Models (PLMs) have achieved substantial advancements in the field of natural language processing. However, in real-world scenarios, data labels are often noisy due to the complex annotation process, making it essential to develop strategies for fine-tuning PLMs with such noisy labels. To this end, we introduce an innovative approach for fine-tuning PLMs using noisy labels, which incorporates the guidance of Large Language Models (LLMs) like ChatGPT. This guidance assists in accurately distinguishing between clean and noisy samples and provides supplementary information beyond the noisy labels, thereby boosting the learning process during fine-tuning PLMs. Extensive experiments on synthetic and real-world noisy datasets further demonstrate the superior advantages of our framework over the state-of-the-art baselines.", }

Adopting a two-stage paradigm of pretraining followed by fine-tuning, Pretrained Language Models (PLMs) have achieved substantial advancements in the field of natural language processing. However, in real-world scenarios, data labels are often noisy due to the complex annotation process, making it essential to develop strategies for fine-tuning PLMs with such noisy labels. To this end, we introduce an innovative approach for fine-tuning PLMs using noisy labels, which incorporates the guidance of Large Language Models (LLMs) like ChatGPT. This guidance assists in accurately distinguishing between clean and noisy samples and provides supplementary information beyond the noisy labels, thereby boosting the learning process during fine-tuning PLMs. Extensive experiments on synthetic and real-world noisy datasets further demonstrate the superior advantages of our framework over the state-of-the-art baselines.

Noise-Robust Fine-Tuning of Pretrained Language Models via External Guidance

findings-emnlp.834

2311.01108

Poster

https://aclanthology.org/2023.findings-emnlp.835.bib

https://aclanthology.org/2023.findings-emnlp.835/

@inproceedings{cao-etal-2023-probabilistic, title = "Probabilistic Tree-of-thought Reasoning for Answering Knowledge-intensive Complex Questions", author = "Cao, Shulin and Zhang, Jiajie and Shi, Jiaxin and Lv, Xin and Yao, Zijun and Tian, Qi and Hou, Lei and Li, Juanzi", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.835", doi = "10.18653/v1/2023.findings-emnlp.835", pages = "12541--12560", abstract = "Large language models (LLMs) are capable of answering knowledge-intensive complex questions with chain-of-thought (CoT) reasoning. However, they tend to generate factually incorrect reasoning steps when the required knowledge is not available or up-to-date in models{'} parameters. Recent works turn to retrieving external knowledge to augment CoT reasoning. Despite being promising, these chain-based methods suffer from: 1) Negative retrieval. Unnecessary or incorrect retrieval may mislead the reasoning; 2) Limited sight. Lacking the ability to look backward or forward, a local error in one step will propagate along the chain. In this paper, we propose a novel approach: Probabilistic Tree-of-thought Reasoning (ProbTree). First, LLMs translate a complex question into a query tree, in which each non-root node denotes a sub-question of its parent node. Then, probabilistic reasoning is conducted over the tree, by solving questions from leaf to root considering the confidence of both question decomposing and answering. During reasoning, for leaf nodes, LLMs choose a more confident answer from Closed-book QA that employs parametric knowledge and Open-book QA that employs retrieved external knowledge, thus eliminating the negative retrieval problem. For non-leaf nodes, with the hierarchical structure, LLMs have broader sights and are able to globally reason with the information from child nodes, thus recovering from local errors. The experiments on three Complex QA datasets under the open-domain setting show that our approach outperforms SOTA methods significantly, demonstrating the effect of probabilistic tree-of-thought reasoning.", }

Large language models (LLMs) are capable of answering knowledge-intensive complex questions with chain-of-thought (CoT) reasoning. However, they tend to generate factually incorrect reasoning steps when the required knowledge is not available or up-to-date in models{'} parameters. Recent works turn to retrieving external knowledge to augment CoT reasoning. Despite being promising, these chain-based methods suffer from: 1) Negative retrieval. Unnecessary or incorrect retrieval may mislead the reasoning; 2) Limited sight. Lacking the ability to look backward or forward, a local error in one step will propagate along the chain. In this paper, we propose a novel approach: Probabilistic Tree-of-thought Reasoning (ProbTree). First, LLMs translate a complex question into a query tree, in which each non-root node denotes a sub-question of its parent node. Then, probabilistic reasoning is conducted over the tree, by solving questions from leaf to root considering the confidence of both question decomposing and answering. During reasoning, for leaf nodes, LLMs choose a more confident answer from Closed-book QA that employs parametric knowledge and Open-book QA that employs retrieved external knowledge, thus eliminating the negative retrieval problem. For non-leaf nodes, with the hierarchical structure, LLMs have broader sights and are able to globally reason with the information from child nodes, thus recovering from local errors. The experiments on three Complex QA datasets under the open-domain setting show that our approach outperforms SOTA methods significantly, demonstrating the effect of probabilistic tree-of-thought reasoning.

Probabilistic Tree-of-thought Reasoning for Answering Knowledge-intensive Complex Questions

findings-emnlp.835

2311.13982

Poster

https://aclanthology.org/2023.findings-emnlp.836.bib

https://aclanthology.org/2023.findings-emnlp.836/

@inproceedings{lee-etal-2023-ensemble, title = "Ensemble-Instruct: Instruction Tuning Data Generation with a Heterogeneous Mixture of {LM}s", author = "Lee, Young-Suk and Sultan, Md and El-Kurdi, Yousef and Naseem, Tahira and Munawar, Asim and Florian, Radu and Roukos, Salim and Astudillo, Ram{\'o}n", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.836", doi = "10.18653/v1/2023.findings-emnlp.836", pages = "12561--12571", abstract = "Using in-context learning (ICL) for data generation, techniques such as Self-Instruct (Wang et al., 2023) or the follow-up Alpaca (Taori et al., 2023) can train strong conversational agents with only a small amount of human supervision. One limitation of these approaches is that they resort to very large language models (around 175B parameters) that are also proprietary and non-public. Here we explore the application of such techniques to language models that are much smaller (around 10B{--}40B parameters) and have permissive licenses. We find the Self-Instruct approach to be less effective at these sizes and propose new ICL methods that draw on two main ideas: (a) categorization and simplification of the ICL templates to make prompt learning easier for the LM, and (b) ensembling over multiple LM outputs to help select high-quality synthetic examples. Our algorithm leverages the 175 Self-Instruct seed tasks and employs separate pipelines for instructions that require an input and instructions that do not. Empirical investigations with different LMs show that: (1) Our proposed method yields higher-quality instruction tuning data than Self-Instruct, (2) It improves performances of both vanilla and instruction-tuned LMs by significant margins, and (3) Smaller instruction-tuned LMs generate more useful examples than their larger un-tuned counterparts.", }

Using in-context learning (ICL) for data generation, techniques such as Self-Instruct (Wang et al., 2023) or the follow-up Alpaca (Taori et al., 2023) can train strong conversational agents with only a small amount of human supervision. One limitation of these approaches is that they resort to very large language models (around 175B parameters) that are also proprietary and non-public. Here we explore the application of such techniques to language models that are much smaller (around 10B{--}40B parameters) and have permissive licenses. We find the Self-Instruct approach to be less effective at these sizes and propose new ICL methods that draw on two main ideas: (a) categorization and simplification of the ICL templates to make prompt learning easier for the LM, and (b) ensembling over multiple LM outputs to help select high-quality synthetic examples. Our algorithm leverages the 175 Self-Instruct seed tasks and employs separate pipelines for instructions that require an input and instructions that do not. Empirical investigations with different LMs show that: (1) Our proposed method yields higher-quality instruction tuning data than Self-Instruct, (2) It improves performances of both vanilla and instruction-tuned LMs by significant margins, and (3) Smaller instruction-tuned LMs generate more useful examples than their larger un-tuned counterparts.

Ensemble-Instruct: Instruction Tuning Data Generation with a Heterogeneous Mixture of LMs

findings-emnlp.836

Poster

https://aclanthology.org/2023.findings-emnlp.837.bib

https://aclanthology.org/2023.findings-emnlp.837/

@inproceedings{nigatu-etal-2023-less, title = "The Less the Merrier? Investigating Language Representation in Multilingual Models", author = "Nigatu, Hellina and Tonja, Atnafu and Kalita, Jugal", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.837", doi = "10.18653/v1/2023.findings-emnlp.837", pages = "12572--12589", abstract = "Multilingual Language Models offer a way to incorporate multiple languages in one model and utilize cross-language transfer learning to improve performance for different Natural Language Processing (NLP) tasks. Despite progress in multilingual models, not all languages are supported as well, particularly in low-resource settings. In this work, we investigate the linguistic representation of different languages in multilingual models. We start by asking the question which languages are supported in popular multilingual models and which languages are left behind. Then, for included languages, we look at models{'} learned representations based on language family and dialect and try to understand how models{'} learned representations for (1) seen and (2) unseen languages vary across different language groups. In addition, we test and analyze performance on downstream tasks such as text generation and Named Entity Recognition. We observe from our experiments that community-centered models{---}models that focus on languages of a given family or geographical location and are built by communities who speak them{---}perform better at distinguishing between languages in the same family for low-resource languages. Our paper contributes to the literature in understanding multilingual models and their shortcomings and offers insights on potential ways to improve them.", }

Multilingual Language Models offer a way to incorporate multiple languages in one model and utilize cross-language transfer learning to improve performance for different Natural Language Processing (NLP) tasks. Despite progress in multilingual models, not all languages are supported as well, particularly in low-resource settings. In this work, we investigate the linguistic representation of different languages in multilingual models. We start by asking the question which languages are supported in popular multilingual models and which languages are left behind. Then, for included languages, we look at models{'} learned representations based on language family and dialect and try to understand how models{'} learned representations for (1) seen and (2) unseen languages vary across different language groups. In addition, we test and analyze performance on downstream tasks such as text generation and Named Entity Recognition. We observe from our experiments that community-centered models{---}models that focus on languages of a given family or geographical location and are built by communities who speak them{---}perform better at distinguishing between languages in the same family for low-resource languages. Our paper contributes to the literature in understanding multilingual models and their shortcomings and offers insights on potential ways to improve them.

The Less the Merrier? Investigating Language Representation in Multilingual Models

findings-emnlp.837

2310.13228

Poster

https://aclanthology.org/2023.findings-emnlp.838.bib

https://aclanthology.org/2023.findings-emnlp.838/

@inproceedings{antypas-etal-2023-supertweeteval, title = "{S}uper{T}weet{E}val: A Challenging, Unified and Heterogeneous Benchmark for Social Media {NLP} Research", author = "Antypas, Dimosthenis and Ushio, Asahi and Barbieri, Francesco and Neves, Leonardo and Rezaee, Kiamehr and Espinosa-Anke, Luis and Pei, Jiaxin and Camacho-Collados, Jose", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.838", doi = "10.18653/v1/2023.findings-emnlp.838", pages = "12590--12607", abstract = "Despite its relevance, the maturity of NLP for social media pales in comparison with general-purpose models, metrics and benchmarks. This fragmented landscape makes it hard for the community to know, for instance, given a task, which is the best performing model and how it compares with others. To alleviate this issue, we introduce a unified benchmark for NLP evaluation in social media, SuperTweetEval, which includes a heterogeneous set of tasks and datasets combined, adapted and constructed from scratch. We benchmarked the performance of a wide range of models on SuperTweetEval and our results suggest that, despite the recent advances in language modelling, social media remains challenging.", }

Despite its relevance, the maturity of NLP for social media pales in comparison with general-purpose models, metrics and benchmarks. This fragmented landscape makes it hard for the community to know, for instance, given a task, which is the best performing model and how it compares with others. To alleviate this issue, we introduce a unified benchmark for NLP evaluation in social media, SuperTweetEval, which includes a heterogeneous set of tasks and datasets combined, adapted and constructed from scratch. We benchmarked the performance of a wide range of models on SuperTweetEval and our results suggest that, despite the recent advances in language modelling, social media remains challenging.

SuperTweetEval: A Challenging, Unified and Heterogeneous Benchmark for Social Media NLP Research

findings-emnlp.838

2310.14757

https://huggingface.co/papers/2310.14757

Poster

https://aclanthology.org/2023.findings-emnlp.839.bib

https://aclanthology.org/2023.findings-emnlp.839/

@inproceedings{fu-etal-2023-enabling, title = "Enabling Unsupervised Neural Machine Translation with Word-level Visual Representations", author = "Fu, Chengpeng and Feng, Xiaocheng and Huang, Yichong and Huo, Wenshuai and Wang, Hui and Qin, Bing and Liu, Ting", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.839", doi = "10.18653/v1/2023.findings-emnlp.839", pages = "12608--12618", abstract = "Unsupervised neural machine translation has recently made remarkable strides, achieving impressive results with the exclusive use of monolingual corpora. Nonetheless, these methods still exhibit fundamental flaws, such as confusing similar words. A straightforward remedy to rectify this drawback is to employ bilingual dictionaries, however, high-quality bilingual dictionaries can be costly to obtain. To overcome this limitation, we propose a method that incorporates images at the word level to augment the lexical mappings. Specifically, our method inserts visual representations into the model, modifying the corresponding embedding layer information. Besides, a visible matrix is adopted to isolate the impact of images on other unrelated words. Experiments on the Multi30k dataset with over 300,000 self-collected images validate the effectiveness in generating more accurate word translation, achieving an improvement of up to $+$2.81 BLEU score, which is comparable or even superior to using bilingual dictionaries.", }

Unsupervised neural machine translation has recently made remarkable strides, achieving impressive results with the exclusive use of monolingual corpora. Nonetheless, these methods still exhibit fundamental flaws, such as confusing similar words. A straightforward remedy to rectify this drawback is to employ bilingual dictionaries, however, high-quality bilingual dictionaries can be costly to obtain. To overcome this limitation, we propose a method that incorporates images at the word level to augment the lexical mappings. Specifically, our method inserts visual representations into the model, modifying the corresponding embedding layer information. Besides, a visible matrix is adopted to isolate the impact of images on other unrelated words. Experiments on the Multi30k dataset with over 300,000 self-collected images validate the effectiveness in generating more accurate word translation, achieving an improvement of up to $+$2.81 BLEU score, which is comparable or even superior to using bilingual dictionaries.

Enabling Unsupervised Neural Machine Translation with Word-level Visual Representations

findings-emnlp.839

Poster

https://aclanthology.org/2023.findings-emnlp.840.bib

https://aclanthology.org/2023.findings-emnlp.840/

@inproceedings{fried-etal-2023-pragmatics, title = "Pragmatics in Language Grounding: Phenomena, Tasks, and Modeling Approaches", author = "Fried, Daniel and Tomlin, Nicholas and Hu, Jennifer and Patel, Roma and Nematzadeh, Aida", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.840", doi = "10.18653/v1/2023.findings-emnlp.840", pages = "12619--12640", abstract = "People rely heavily on context to enrich meaning beyond what is literally said, enabling concise but effective communication. To interact successfully and naturally with people, user-facing artificial intelligence systems will require similar skills in pragmatics: relying on various types of context {---} from shared linguistic goals and conventions, to the visual and embodied world {---} to use language effectively. We survey existing grounded settings and pragmatic modeling approaches and analyze how the task goals, environmental contexts, and communicative affordances in each work enrich linguistic meaning. We present recommendations for future grounded task design to naturally elicit pragmatic phenomena, and suggest directions that focus on a broader range of communicative contexts and affordances.", }

People rely heavily on context to enrich meaning beyond what is literally said, enabling concise but effective communication. To interact successfully and naturally with people, user-facing artificial intelligence systems will require similar skills in pragmatics: relying on various types of context {---} from shared linguistic goals and conventions, to the visual and embodied world {---} to use language effectively. We survey existing grounded settings and pragmatic modeling approaches and analyze how the task goals, environmental contexts, and communicative affordances in each work enrich linguistic meaning. We present recommendations for future grounded task design to naturally elicit pragmatic phenomena, and suggest directions that focus on a broader range of communicative contexts and affordances.

Pragmatics in Language Grounding: Phenomena, Tasks, and Modeling Approaches

findings-emnlp.840

2211.08371

Poster

https://aclanthology.org/2023.findings-emnlp.841.bib

https://aclanthology.org/2023.findings-emnlp.841/

@inproceedings{pham-etal-2023-misca, title = "{MISCA}: A Joint Model for Multiple Intent Detection and Slot Filling with Intent-Slot Co-Attention", author = "Pham, Thinh and Tran, Chi and Nguyen, Dat Quoc", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.841", doi = "10.18653/v1/2023.findings-emnlp.841", pages = "12641--12650", abstract = "The research study of detecting multiple intents and filling slots is becoming more popular because of its relevance to complicated real-world situations. Recent advanced approaches, which are joint models based on graphs, might still face two potential issues: (i) the uncertainty introduced by constructing graphs based on preliminary intents and slots, which may transfer intent-slot correlation information to incorrect label node destinations, and (ii) direct incorporation of multiple intent labels for each token w.r.t. token-level intent voting might potentially lead to incorrect slot predictions, thereby hurting the overall performance. To address these two issues, we propose a joint model named MISCA. Our MISCA introduces an intent-slot co-attention mechanism and an underlying layer of label attention mechanism. These mechanisms enable MISCA to effectively capture correlations between intents and slot labels, eliminating the need for graph construction. They also facilitate the transfer of correlation information in both directions: from intents to slots and from slots to intents, through multiple levels of label-specific representations, without relying on token-level intent information. Experimental results show that MISCA outperforms previous models, achieving new state-of-the-art overall accuracy performances on two benchmark datasets MixATIS and MixSNIPS. This highlights the effectiveness of our attention mechanisms.", }

The research study of detecting multiple intents and filling slots is becoming more popular because of its relevance to complicated real-world situations. Recent advanced approaches, which are joint models based on graphs, might still face two potential issues: (i) the uncertainty introduced by constructing graphs based on preliminary intents and slots, which may transfer intent-slot correlation information to incorrect label node destinations, and (ii) direct incorporation of multiple intent labels for each token w.r.t. token-level intent voting might potentially lead to incorrect slot predictions, thereby hurting the overall performance. To address these two issues, we propose a joint model named MISCA. Our MISCA introduces an intent-slot co-attention mechanism and an underlying layer of label attention mechanism. These mechanisms enable MISCA to effectively capture correlations between intents and slot labels, eliminating the need for graph construction. They also facilitate the transfer of correlation information in both directions: from intents to slots and from slots to intents, through multiple levels of label-specific representations, without relying on token-level intent information. Experimental results show that MISCA outperforms previous models, achieving new state-of-the-art overall accuracy performances on two benchmark datasets MixATIS and MixSNIPS. This highlights the effectiveness of our attention mechanisms.

MISCA: A Joint Model for Multiple Intent Detection and Slot Filling with Intent-Slot Co-Attention

findings-emnlp.841

2312.05741

Poster

https://aclanthology.org/2023.findings-emnlp.842.bib

https://aclanthology.org/2023.findings-emnlp.842/

@inproceedings{hou-etal-2023-enhancing, title = "Enhancing Emotion Recognition in Conversation via Multi-view Feature Alignment and Memorization", author = "Hou, Guiyang and Shen, Yongliang and Zhang, Wenqi and Xue, Wei and Lu, Weiming", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.842", doi = "10.18653/v1/2023.findings-emnlp.842", pages = "12651--12663", abstract = "Emotion recognition in conversation (ERC) has attracted increasing attention in natural language processing community. Previous work commonly first extract semantic-view features via fine-tuning PLMs, then models context-view features based on the obtained semantic-view features by various graph neural networks. However, it is difficult to fully model interaction between utterances simply through a graph neural network and the features at semantic-view and context-view are not well aligned. Moreover, the previous parametric learning paradigm struggle to learn the patterns of tail class given fewer instances. To this end, we treat the pre-trained conversation model as a prior knowledge base and from which we elicit correlations between utterances by a probing procedure. And we adopt supervised contrastive learning to align semantic-view and context-view features, these two views of features work together in a complementary manner, contributing to ERC from distinct perspectives. Meanwhile, we propose a new semi-parametric paradigm of inferencing through memorization to solve the recognition problem of tail class samples. We consistently achieve state-of-the-art results on four widely used benchmarks. Extensive experiments demonstrate the effectiveness of our proposed multi-view feature alignment and memorization.", }

Emotion recognition in conversation (ERC) has attracted increasing attention in natural language processing community. Previous work commonly first extract semantic-view features via fine-tuning PLMs, then models context-view features based on the obtained semantic-view features by various graph neural networks. However, it is difficult to fully model interaction between utterances simply through a graph neural network and the features at semantic-view and context-view are not well aligned. Moreover, the previous parametric learning paradigm struggle to learn the patterns of tail class given fewer instances. To this end, we treat the pre-trained conversation model as a prior knowledge base and from which we elicit correlations between utterances by a probing procedure. And we adopt supervised contrastive learning to align semantic-view and context-view features, these two views of features work together in a complementary manner, contributing to ERC from distinct perspectives. Meanwhile, we propose a new semi-parametric paradigm of inferencing through memorization to solve the recognition problem of tail class samples. We consistently achieve state-of-the-art results on four widely used benchmarks. Extensive experiments demonstrate the effectiveness of our proposed multi-view feature alignment and memorization.

Enhancing Emotion Recognition in Conversation via Multi-view Feature Alignment and Memorization

findings-emnlp.842

Poster

https://aclanthology.org/2023.findings-emnlp.843.bib

https://aclanthology.org/2023.findings-emnlp.843/

@inproceedings{wang-walther-2023-mandarin, title = "{M}andarin classifier systems optimize to accommodate communicative pressures", author = "Wang, Yamei and Walther, G{\'e}raldine", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.843", doi = "10.18653/v1/2023.findings-emnlp.843", pages = "12664--12674", abstract = "Previous work on noun classification implies that gender systems are inherently optimized to accommodate communicative pressures on human language learning and processing (Dye. et al 2017, 2018). They state that languages make use of either grammatical (e.g., gender) or probabilistic (pre-nominal modifiers) to smoothe the entropy of nouns in context. We show that even languages that are considered genderless, like Mandarin Chinese, possess a noun classification device that plays the same functional role as gender markers. Based on close to 1M Mandarin noun phrases extracted from the Leipzig Corpora Collection (Goldhahn et al. 2012) and their corresponding fastText embeddings (Bojanowski et al. 2016), we show that noun-classifier combinations are sensitive to same frequency, similarity, and co-occurrence interactions that structure gender systems. We also present the first study of the effects of the interaction between grammatical and probabilisitic noun classification.", }

Previous work on noun classification implies that gender systems are inherently optimized to accommodate communicative pressures on human language learning and processing (Dye. et al 2017, 2018). They state that languages make use of either grammatical (e.g., gender) or probabilistic (pre-nominal modifiers) to smoothe the entropy of nouns in context. We show that even languages that are considered genderless, like Mandarin Chinese, possess a noun classification device that plays the same functional role as gender markers. Based on close to 1M Mandarin noun phrases extracted from the Leipzig Corpora Collection (Goldhahn et al. 2012) and their corresponding fastText embeddings (Bojanowski et al. 2016), we show that noun-classifier combinations are sensitive to same frequency, similarity, and co-occurrence interactions that structure gender systems. We also present the first study of the effects of the interaction between grammatical and probabilisitic noun classification.

Mandarin classifier systems optimize to accommodate communicative pressures

findings-emnlp.843

Poster

https://aclanthology.org/2023.findings-emnlp.844.bib

https://aclanthology.org/2023.findings-emnlp.844/

@inproceedings{wang-etal-2023-probing, title = "Probing Representations for Document-level Event Extraction", author = "Wang, Barry and Du, Xinya and Cardie, Claire", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.844", doi = "10.18653/v1/2023.findings-emnlp.844", pages = "12675--12683", abstract = "The probing classifiers framework has been employed for interpreting deep neural network models for a variety of natural language processing (NLP) applications. Studies, however, have largely focused on sentencelevel NLP tasks. This work is the first to apply the probing paradigm to representations learned for document-level information extraction (IE). We designed eight embedding probes to analyze surface, semantic, and event-understanding capabilities relevant to document-level event extraction. We apply them to the representations acquired by learning models from three different LLM-based document-level IE approaches on a standard dataset. We found that trained encoders from these models yield embeddings that can modestly improve argument detections and labeling but only slightly enhance event-level tasks, albeit trade-offs in information helpful for coherence and event-type prediction. We further found that encoder models struggle with document length and cross-sentence discourse.", }

The probing classifiers framework has been employed for interpreting deep neural network models for a variety of natural language processing (NLP) applications. Studies, however, have largely focused on sentencelevel NLP tasks. This work is the first to apply the probing paradigm to representations learned for document-level information extraction (IE). We designed eight embedding probes to analyze surface, semantic, and event-understanding capabilities relevant to document-level event extraction. We apply them to the representations acquired by learning models from three different LLM-based document-level IE approaches on a standard dataset. We found that trained encoders from these models yield embeddings that can modestly improve argument detections and labeling but only slightly enhance event-level tasks, albeit trade-offs in information helpful for coherence and event-type prediction. We further found that encoder models struggle with document length and cross-sentence discourse.

Probing Representations for Document-level Event Extraction

findings-emnlp.844

2310.15316

Poster

https://aclanthology.org/2023.findings-emnlp.845.bib

https://aclanthology.org/2023.findings-emnlp.845/

@inproceedings{zhou-etal-2023-cultural, title = "Cultural Compass: Predicting Transfer Learning Success in Offensive Language Detection with Cultural Features", author = "Zhou, Li and Karamolegkou, Antonia and Chen, Wenyu and Hershcovich, Daniel", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.845", doi = "10.18653/v1/2023.findings-emnlp.845", pages = "12684--12702", abstract = "The increasing ubiquity of language technology necessitates a shift towards considering cultural diversity in the machine learning realm, particularly for subjective tasks that rely heavily on cultural nuances, such as Offensive Language Detection (OLD). Current understanding underscores that these tasks are substantially influenced by cultural values, however, a notable gap exists in determining if cultural features can accurately predict the success of cross-cultural transfer learning for such subjective tasks. Addressing this, our study delves into the intersection of cultural features and transfer learning effectiveness. The findings reveal that cultural value surveys indeed possess a predictive power for cross-cultural transfer learning success in OLD tasks, and that it can be further improved using offensive word distance. Based on these results, we advocate for the integration of cultural information into datasets. Additionally, we recommend leveraging data sources rich in cultural information, such as surveys, to enhance cultural adaptability. Our research signifies a step forward in the quest for more inclusive, culturally sensitive language technologies.", }

The increasing ubiquity of language technology necessitates a shift towards considering cultural diversity in the machine learning realm, particularly for subjective tasks that rely heavily on cultural nuances, such as Offensive Language Detection (OLD). Current understanding underscores that these tasks are substantially influenced by cultural values, however, a notable gap exists in determining if cultural features can accurately predict the success of cross-cultural transfer learning for such subjective tasks. Addressing this, our study delves into the intersection of cultural features and transfer learning effectiveness. The findings reveal that cultural value surveys indeed possess a predictive power for cross-cultural transfer learning success in OLD tasks, and that it can be further improved using offensive word distance. Based on these results, we advocate for the integration of cultural information into datasets. Additionally, we recommend leveraging data sources rich in cultural information, such as surveys, to enhance cultural adaptability. Our research signifies a step forward in the quest for more inclusive, culturally sensitive language technologies.

Cultural Compass: Predicting Transfer Learning Success in Offensive Language Detection with Cultural Features

findings-emnlp.845

2310.06458

Poster

https://aclanthology.org/2023.findings-emnlp.846.bib

https://aclanthology.org/2023.findings-emnlp.846/

@inproceedings{moryossef-etal-2023-linguistically, title = "Linguistically Motivated Sign Language Segmentation", author = {Moryossef, Amit and Jiang, Zifan and M{\"u}ller, Mathias and Ebling, Sarah and Goldberg, Yoav}, editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.846", doi = "10.18653/v1/2023.findings-emnlp.846", pages = "12703--12724", abstract = "Sign language segmentation is a crucial task in sign language processing systems. It enables downstream tasks such as sign recognition, transcription, and machine translation. In this work, we consider two kinds of segmentation: segmentation into individual signs and segmentation into \textit{phrases}, larger units comprising several signs. We propose a novel approach to jointly model these two tasks. Our method is motivated by linguistic cues observed in sign language corpora. We replace the predominant IO tagging scheme with BIO tagging to account for continuous signing. Given that prosody plays a significant role in phrase boundaries, we explore the use of optical flow features. We also provide an extensive analysis of hand shapes and 3D hand normalization. We find that introducing BIO tagging is necessary to model sign boundaries. Explicitly encoding prosody by optical flow improves segmentation in shallow models, but its contribution is negligible in deeper models. Careful tuning of the decoding algorithm atop the models further improves the segmentation quality. We demonstrate that our final models generalize to out-of-domain video content in a different signed language, even under a zero-shot setting. We observe that including optical flow and 3D hand normalization enhances the robustness of the model in this context.", }

Sign language segmentation is a crucial task in sign language processing systems. It enables downstream tasks such as sign recognition, transcription, and machine translation. In this work, we consider two kinds of segmentation: segmentation into individual signs and segmentation into \textit{phrases}, larger units comprising several signs. We propose a novel approach to jointly model these two tasks. Our method is motivated by linguistic cues observed in sign language corpora. We replace the predominant IO tagging scheme with BIO tagging to account for continuous signing. Given that prosody plays a significant role in phrase boundaries, we explore the use of optical flow features. We also provide an extensive analysis of hand shapes and 3D hand normalization. We find that introducing BIO tagging is necessary to model sign boundaries. Explicitly encoding prosody by optical flow improves segmentation in shallow models, but its contribution is negligible in deeper models. Careful tuning of the decoding algorithm atop the models further improves the segmentation quality. We demonstrate that our final models generalize to out-of-domain video content in a different signed language, even under a zero-shot setting. We observe that including optical flow and 3D hand normalization enhances the robustness of the model in this context.

Linguistically Motivated Sign Language Segmentation

findings-emnlp.846

2310.13960

Poster

https://aclanthology.org/2023.findings-emnlp.847.bib

https://aclanthology.org/2023.findings-emnlp.847/

@inproceedings{luo-etal-2023-weighting, title = "Re-weighting Tokens: A Simple and Effective Active Learning Strategy for Named Entity Recognition", author = "Luo, Haocheng and Tan, Wei and Nguyen, Ngoc and Du, Lan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.847", doi = "10.18653/v1/2023.findings-emnlp.847", pages = "12725--12734", abstract = "Active learning, a widely adopted technique for enhancing machine learning models in text and image classification tasks with limited annotation resources, has received relatively little attention in the domain of Named Entity Recognition (NER). The challenge of data imbalance in NER has hindered the effectiveness of active learning, as sequence labellers lack sufficient learning signals. To address these challenges, this paper presents a novel re-weighting-based active learning strategy that assigns dynamic smoothing weights to individual tokens. This adaptable strategy is compatible with various token-level acquisition functions and contributes to the development of robust active learners. Experimental results on multiple corpora demonstrate the substantial performance improvement achieved by incorporating our re-weighting strategy into existing acquisition functions, validating its practical efficacy. We will release our implementation upon the publication of this paper.", }

Active learning, a widely adopted technique for enhancing machine learning models in text and image classification tasks with limited annotation resources, has received relatively little attention in the domain of Named Entity Recognition (NER). The challenge of data imbalance in NER has hindered the effectiveness of active learning, as sequence labellers lack sufficient learning signals. To address these challenges, this paper presents a novel re-weighting-based active learning strategy that assigns dynamic smoothing weights to individual tokens. This adaptable strategy is compatible with various token-level acquisition functions and contributes to the development of robust active learners. Experimental results on multiple corpora demonstrate the substantial performance improvement achieved by incorporating our re-weighting strategy into existing acquisition functions, validating its practical efficacy. We will release our implementation upon the publication of this paper.

Re-weighting Tokens: A Simple and Effective Active Learning Strategy for Named Entity Recognition

findings-emnlp.847

2311.00906

Poster

https://aclanthology.org/2023.findings-emnlp.848.bib

https://aclanthology.org/2023.findings-emnlp.848/

@inproceedings{koksal-etal-2023-language, title = "Language-Agnostic Bias Detection in Language Models with Bias Probing", author = {K{\"o}ksal, Abdullatif and Yalcin, Omer and Akbiyik, Ahmet and Kilavuz, M. and Korhonen, Anna and Schuetze, Hinrich}, editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.848", doi = "10.18653/v1/2023.findings-emnlp.848", pages = "12735--12747", abstract = "Pretrained language models (PLMs) are key components in NLP, but they contain strong social biases. Quantifying these biases is challenging because current methods focusing on fill-the-mask objectives are sensitive to slight changes in input. To address this, we propose a bias probing technique called LABDet, for evaluating social bias in PLMs with a robust and language-agnostic method. For nationality as a case study, we show that LABDet {``}surfaces{''} nationality bias by training a classifier on top of a frozen PLM on non-nationality sentiment detection. We find consistent patterns of nationality bias across monolingual PLMs in six languages that align with historical and political context. We also show for English BERT that bias surfaced by LABDet correlates well with bias in the pretraining data; thus, our work is one of the few studies that directly links pretraining data to PLM behavior. Finally, we verify LABDet{'}s reliability and applicability to different templates and languages through an extensive set of robustness checks. We publicly share our code and dataset in https://github.com/akoksal/LABDet.", }

Pretrained language models (PLMs) are key components in NLP, but they contain strong social biases. Quantifying these biases is challenging because current methods focusing on fill-the-mask objectives are sensitive to slight changes in input. To address this, we propose a bias probing technique called LABDet, for evaluating social bias in PLMs with a robust and language-agnostic method. For nationality as a case study, we show that LABDet {``}surfaces{''} nationality bias by training a classifier on top of a frozen PLM on non-nationality sentiment detection. We find consistent patterns of nationality bias across monolingual PLMs in six languages that align with historical and political context. We also show for English BERT that bias surfaced by LABDet correlates well with bias in the pretraining data; thus, our work is one of the few studies that directly links pretraining data to PLM behavior. Finally, we verify LABDet{'}s reliability and applicability to different templates and languages through an extensive set of robustness checks. We publicly share our code and dataset in https://github.com/akoksal/LABDet.

Language-Agnostic Bias Detection in Language Models with Bias Probing

findings-emnlp.848

2305.13302

Poster

https://aclanthology.org/2023.findings-emnlp.849.bib

https://aclanthology.org/2023.findings-emnlp.849/

@inproceedings{yu-etal-2023-compleqa, title = "{C}omple{QA}: Benchmarking the Impacts of Knowledge Graph Completion Methods on Question Answering", author = "Yu, Donghan and Gu, Yu and Xiong, Chenyan and Yang, Yiming", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.849", doi = "10.18653/v1/2023.findings-emnlp.849", pages = "12748--12755", abstract = "How much success in Knowledge Graph Completion (KGC) would translate into the performance enhancement in downstream tasks is an important question that has not been studied in depth. In this paper, we introduce a novel benchmark, namely CompleQA, to comprehensively assess the influence of representative KGC methods on Knowledge Graph Question Answering (KGQA), one of the most important downstream applications. This benchmark includes a knowledge graph with 3 million triplets across 5 distinct domains, coupled with over 5000 question-answering pairs and a completion dataset that is well-aligned with these questions. Our evaluation of four well-known KGC methods in combination with two state-of-the-art KGQA systems shows that effective KGC can significantly mitigate the impact of knowledge graph incompleteness on question-answering performance. Surprisingly, we also find that the best-performing KGC method(s) does not necessarily lead to the best QA results, underscoring the need to consider downstream applications when doing KGC.", }

How much success in Knowledge Graph Completion (KGC) would translate into the performance enhancement in downstream tasks is an important question that has not been studied in depth. In this paper, we introduce a novel benchmark, namely CompleQA, to comprehensively assess the influence of representative KGC methods on Knowledge Graph Question Answering (KGQA), one of the most important downstream applications. This benchmark includes a knowledge graph with 3 million triplets across 5 distinct domains, coupled with over 5000 question-answering pairs and a completion dataset that is well-aligned with these questions. Our evaluation of four well-known KGC methods in combination with two state-of-the-art KGQA systems shows that effective KGC can significantly mitigate the impact of knowledge graph incompleteness on question-answering performance. Surprisingly, we also find that the best-performing KGC method(s) does not necessarily lead to the best QA results, underscoring the need to consider downstream applications when doing KGC.

CompleQA: Benchmarking the Impacts of Knowledge Graph Completion Methods on Question Answering

findings-emnlp.849

Poster

https://aclanthology.org/2023.findings-emnlp.850.bib

https://aclanthology.org/2023.findings-emnlp.850/

@inproceedings{lin-etal-2023-improving, title = "Improving Multi-Criteria {C}hinese Word Segmentation through Learning Sentence Representation", author = "Lin, Chun and Lin, Ying-Jia and Yeh, Chia-Jen and Li, Yi-Ting and Yang, Ching-Wen and Kao, Hung-Yu", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.850", doi = "10.18653/v1/2023.findings-emnlp.850", pages = "12756--12763", abstract = "Recent Chinese word segmentation (CWS) models have shown competitive performance with pre-trained language models{'} knowledge. However, these models tend to learn the segmentation knowledge through in-vocabulary words rather than understanding the meaning of the entire context. To address this issue, we introduce a context-aware approach that incorporates unsupervised sentence representation learning over different dropout masks into the multi-criteria training framework. We demonstrate that our approach reaches state-of-the-art (SoTA) performance on F1 scores for six of the nine CWS benchmark datasets and out-of-vocabulary (OOV) recalls for eight of nine. Further experiments discover that substantial improvements can be brought with various sentence representation objectives.", }

Recent Chinese word segmentation (CWS) models have shown competitive performance with pre-trained language models{'} knowledge. However, these models tend to learn the segmentation knowledge through in-vocabulary words rather than understanding the meaning of the entire context. To address this issue, we introduce a context-aware approach that incorporates unsupervised sentence representation learning over different dropout masks into the multi-criteria training framework. We demonstrate that our approach reaches state-of-the-art (SoTA) performance on F1 scores for six of the nine CWS benchmark datasets and out-of-vocabulary (OOV) recalls for eight of nine. Further experiments discover that substantial improvements can be brought with various sentence representation objectives.

Improving Multi-Criteria Chinese Word Segmentation through Learning Sentence Representation

findings-emnlp.850

Poster

https://aclanthology.org/2023.findings-emnlp.851.bib

https://aclanthology.org/2023.findings-emnlp.851/

@inproceedings{zhao-etal-2023-joint, title = "A Joint Matrix Factorization Analysis of Multilingual Representations", author = "Zhao, Zheng and Ziser, Yftah and Webber, Bonnie and Cohen, Shay", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.851", doi = "10.18653/v1/2023.findings-emnlp.851", pages = "12764--12783", abstract = "We present an analysis tool based on joint matrix factorization for comparing latent representations of multilingual and monolingual models. An alternative to probing, this tool allows us to analyze multiple sets of representations in a joint manner. Using this tool, we study to what extent and how morphosyntactic features are reflected in the representations learned by multilingual pre-trained models. We conduct a large-scale empirical study of over 33 languages and 17 morphosyntactic categories. Our findings demonstrate variations in the encoding of morphosyntactic information across upper and lower layers, with category-specific differences influenced by language properties. Hierarchical clustering of the factorization outputs yields a tree structure that is related to phylogenetic trees manually crafted by linguists. Moreover, we find the factorization outputs exhibit strong associations with performance observed across different cross-lingual tasks. We release our code to facilitate future research.", }

We present an analysis tool based on joint matrix factorization for comparing latent representations of multilingual and monolingual models. An alternative to probing, this tool allows us to analyze multiple sets of representations in a joint manner. Using this tool, we study to what extent and how morphosyntactic features are reflected in the representations learned by multilingual pre-trained models. We conduct a large-scale empirical study of over 33 languages and 17 morphosyntactic categories. Our findings demonstrate variations in the encoding of morphosyntactic information across upper and lower layers, with category-specific differences influenced by language properties. Hierarchical clustering of the factorization outputs yields a tree structure that is related to phylogenetic trees manually crafted by linguists. Moreover, we find the factorization outputs exhibit strong associations with performance observed across different cross-lingual tasks. We release our code to facilitate future research.

A Joint Matrix Factorization Analysis of Multilingual Representations

findings-emnlp.851

2310.15513

Poster

https://aclanthology.org/2023.findings-emnlp.852.bib

https://aclanthology.org/2023.findings-emnlp.852/

@inproceedings{slobodkin-etal-2023-dont, title = "Don{'}t Add, don{'}t Miss: Effective Content Preserving Generation from Pre-Selected Text Spans", author = "Slobodkin, Aviv and Caciularu, Avi and Hirsch, Eran and Dagan, Ido", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.852", doi = "10.18653/v1/2023.findings-emnlp.852", pages = "12784--12800", abstract = "The recently introduced Controlled Text Reduction (CTR) task isolates the text generation step within typical summarization-style tasks. It does so by challenging models to generate coherent text conforming to pre-selected content within the input text ({``}highlights{''}). This framing enables increased modularity in summarization-like tasks, allowing to couple a single CTR model with various content-selection setups and modules. However, there are currently no reliable CTR models, while the performance of the existing baseline for the task is mediocre, falling short of practical utility. Here, we address this gap by introducing a high-quality, open-source CTR model that tackles two prior key limitations: inadequate enforcement of the content-preservation constraint, and suboptimal silver training data. Addressing these, we amplify the content-preservation constraint in both training, via RL, and inference, via a controlled decoding strategy. Further, we substantially improve the silver training data quality via GPT-4 distillation. Overall, pairing the distilled dataset with the highlight-adherence strategies yields marked gains over the current baseline, of up to 30 ROUGE-L points, providing a reliable CTR model for downstream use.", }

The recently introduced Controlled Text Reduction (CTR) task isolates the text generation step within typical summarization-style tasks. It does so by challenging models to generate coherent text conforming to pre-selected content within the input text ({``}highlights{''}). This framing enables increased modularity in summarization-like tasks, allowing to couple a single CTR model with various content-selection setups and modules. However, there are currently no reliable CTR models, while the performance of the existing baseline for the task is mediocre, falling short of practical utility. Here, we address this gap by introducing a high-quality, open-source CTR model that tackles two prior key limitations: inadequate enforcement of the content-preservation constraint, and suboptimal silver training data. Addressing these, we amplify the content-preservation constraint in both training, via RL, and inference, via a controlled decoding strategy. Further, we substantially improve the silver training data quality via GPT-4 distillation. Overall, pairing the distilled dataset with the highlight-adherence strategies yields marked gains over the current baseline, of up to 30 ROUGE-L points, providing a reliable CTR model for downstream use.

Don't Add, don't Miss: Effective Content Preserving Generation from Pre-Selected Text Spans

findings-emnlp.852

Poster

https://aclanthology.org/2023.findings-emnlp.853.bib

https://aclanthology.org/2023.findings-emnlp.853/

@inproceedings{tambwekar-etal-2023-computational, title = "A Computational Interface to Translate Strategic Intent from Unstructured Language in a Low-Data Setting", author = "Tambwekar, Pradyumna and Dodeja, Lakshita and Vaska, Nathan and Xu, Wei and Gombolay, Matthew", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.853", doi = "10.18653/v1/2023.findings-emnlp.853", pages = "12801--12819", abstract = "Many real-world tasks involve a mixed-initiative setup, wherein humans and AI systems collaboratively perform a task. While significant work has been conducted towards enabling humans to specify, through language, exactly how an agent should complete a task (i.e., low-level specification), prior work lacks on interpreting the high-level strategic intent of the human commanders. Parsing strategic intent from language will allow autonomous systems to independently operate according to the user{'}s plan without frequent guidance or instruction. In this paper, we build a computational interface capable of translating unstructured language strategies into actionable intent in the form of goals and constraints. Leveraging a game environment, we collect a dataset of over 1000 examples, mapping language strategies to the corresponding goals and constraints, and show that our model, trained on this dataset, significantly outperforms human interpreters in inferring strategic intent (i.e., goals and constraints) from language (p {\textless} 0.05). Furthermore, we show that our model (125M parameters) significantly outperforms ChatGPT for this task (p {\textless} 0.05) in a low-data setting.", }

Many real-world tasks involve a mixed-initiative setup, wherein humans and AI systems collaboratively perform a task. While significant work has been conducted towards enabling humans to specify, through language, exactly how an agent should complete a task (i.e., low-level specification), prior work lacks on interpreting the high-level strategic intent of the human commanders. Parsing strategic intent from language will allow autonomous systems to independently operate according to the user{'}s plan without frequent guidance or instruction. In this paper, we build a computational interface capable of translating unstructured language strategies into actionable intent in the form of goals and constraints. Leveraging a game environment, we collect a dataset of over 1000 examples, mapping language strategies to the corresponding goals and constraints, and show that our model, trained on this dataset, significantly outperforms human interpreters in inferring strategic intent (i.e., goals and constraints) from language (p {\textless} 0.05). Furthermore, we show that our model (125M parameters) significantly outperforms ChatGPT for this task (p {\textless} 0.05) in a low-data setting.

A Computational Interface to Translate Strategic Intent from Unstructured Language in a Low-Data Setting

findings-emnlp.853

2208.08374

Poster

https://aclanthology.org/2023.findings-emnlp.854.bib

https://aclanthology.org/2023.findings-emnlp.854/

@inproceedings{li-etal-2023-hfmre, title = "{HFMRE}: Constructing {H}uffman Tree in Bags to Find Excellent Instances for Distantly Supervised Relation Extraction", author = "Li, Min and Shao, Cong and Li, Gang and Zhou, Mingle", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.854", doi = "10.18653/v1/2023.findings-emnlp.854", pages = "12820--12832", abstract = "Since the introduction of distantly supervised relation extraction methods, numerous approaches have been developed, the most representative of which is multi-instance learning (MIL). To find reliable features that are most representative of multi-instance bags, aggregation strategies such as AVG (average), ONE (at least one), and ATT (sentence-level attention) are commonly used. These strategies tend to train third-party vectors to select sentence-level features, leaving it to the third party to decide/identify what is noise, ignoring the intrinsic associations that naturally exist from sentence to sentence. In this paper, we propose the concept of circular cosine similarity, which is used to explicitly show the intrinsic associations between sentences within a bag. We also consider the previous methods to be a crude denoising process as they are interrupted and do not have a continuous noise detection procedure. Following this consideration, we implement a relation extraction framework (HFMRE) that relies on the Huffman tree, where sentences are considered as leaf nodes and circular cosine similarity are considered as node weights. HFMRE can continuously and iteratively discriminate noise and aggregated features during the construction of the Huffman tree, eventually finding an excellent instance that is representative of a bag-level feature. The experiments demonstrate the remarkable effectiveness of our method, outperforming previously advanced baselines on the popular DSRE datasets.", }

Since the introduction of distantly supervised relation extraction methods, numerous approaches have been developed, the most representative of which is multi-instance learning (MIL). To find reliable features that are most representative of multi-instance bags, aggregation strategies such as AVG (average), ONE (at least one), and ATT (sentence-level attention) are commonly used. These strategies tend to train third-party vectors to select sentence-level features, leaving it to the third party to decide/identify what is noise, ignoring the intrinsic associations that naturally exist from sentence to sentence. In this paper, we propose the concept of circular cosine similarity, which is used to explicitly show the intrinsic associations between sentences within a bag. We also consider the previous methods to be a crude denoising process as they are interrupted and do not have a continuous noise detection procedure. Following this consideration, we implement a relation extraction framework (HFMRE) that relies on the Huffman tree, where sentences are considered as leaf nodes and circular cosine similarity are considered as node weights. HFMRE can continuously and iteratively discriminate noise and aggregated features during the construction of the Huffman tree, eventually finding an excellent instance that is representative of a bag-level feature. The experiments demonstrate the remarkable effectiveness of our method, outperforming previously advanced baselines on the popular DSRE datasets.

HFMRE: Constructing Huffman Tree in Bags to Find Excellent Instances for Distantly Supervised Relation Extraction

findings-emnlp.854

Poster

https://aclanthology.org/2023.findings-emnlp.855.bib

https://aclanthology.org/2023.findings-emnlp.855/

@inproceedings{bhat-etal-2023-disco, title = "{DISCO}: A Large Scale Human Annotated Corpus for Disfluency Correction in {I}ndo-{E}uropean Languages", author = "Bhat, Vineet and Jyothi, Preethi and Bhattacharyya, Pushpak", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.855", doi = "10.18653/v1/2023.findings-emnlp.855", pages = "12833--12857", abstract = "Disfluency correction (DC) is the process of removing disfluent elements like fillers, repetitions and corrections from spoken utterances to create readable and interpretable text. DC is a vital post-processing step applied to Automatic Speech Recognition (ASR) outputs, before subsequent processing by downstream language understanding tasks. Existing DC research has primarily focused on English due to the unavailability of large-scale open-source datasets. Towards the goal of multilingual disfluency correction, we present a high-quality human-annotated DC corpus covering four important Indo-European languages: English, Hindi, German and French. We provide extensive analysis of results of state-of-the-art DC models across all four languages obtaining F1 scores of 97.55 (English), 94.29 (Hindi), 95.89 (German) and 92.97 (French). To demonstrate the benefits of DC on downstream tasks, we show that DC leads to 5.65 points increase in BLEU scores on average when used in conjunction with a state-of-the-art Machine Translation (MT) system. We release code to run our experiments along with our annotated dataset here.", }

Disfluency correction (DC) is the process of removing disfluent elements like fillers, repetitions and corrections from spoken utterances to create readable and interpretable text. DC is a vital post-processing step applied to Automatic Speech Recognition (ASR) outputs, before subsequent processing by downstream language understanding tasks. Existing DC research has primarily focused on English due to the unavailability of large-scale open-source datasets. Towards the goal of multilingual disfluency correction, we present a high-quality human-annotated DC corpus covering four important Indo-European languages: English, Hindi, German and French. We provide extensive analysis of results of state-of-the-art DC models across all four languages obtaining F1 scores of 97.55 (English), 94.29 (Hindi), 95.89 (German) and 92.97 (French). To demonstrate the benefits of DC on downstream tasks, we show that DC leads to 5.65 points increase in BLEU scores on average when used in conjunction with a state-of-the-art Machine Translation (MT) system. We release code to run our experiments along with our annotated dataset here.

DISCO: A Large Scale Human Annotated Corpus for Disfluency Correction in Indo-European Languages

findings-emnlp.855

2310.16749

Poster

https://aclanthology.org/2023.findings-emnlp.856.bib

https://aclanthology.org/2023.findings-emnlp.856/

@inproceedings{xu-etal-2023-towards-parameter, title = "Towards Being Parameter-Efficient: A Stratified Sparsely Activated Transformer with Dynamic Capacity", author = "Xu, Haoran and Elbayad, Maha and Murray, Kenton and Maillard, Jean and Goswami, Vedanuj", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.856", doi = "10.18653/v1/2023.findings-emnlp.856", pages = "12858--12870", abstract = "Mixture-of-experts (MoE) models that employ sparse activation have demonstrated effectiveness in significantly increasing the number of parameters while maintaining low computational requirements per token. However, recent studies have established that MoE models are inherently parameter-inefficient as the improvement in performance diminishes with an increasing number of experts. We hypothesize this parameter inefficiency is a result of all experts having equal capacity, which may not adequately meet the varying complexity requirements of different tokens or tasks. In light of this, we propose Stratified Mixture of Experts (SMoE) models, which feature a stratified structure and can assign dynamic capacity to different tokens. We demonstrate the effectiveness of SMoE on three multilingual machine translation benchmarks, containing 4, 15, and 94 language pairs, respectively. We show that SMoE outperforms multiple state-of-the-art MoE models with the same or fewer parameters.", }

Mixture-of-experts (MoE) models that employ sparse activation have demonstrated effectiveness in significantly increasing the number of parameters while maintaining low computational requirements per token. However, recent studies have established that MoE models are inherently parameter-inefficient as the improvement in performance diminishes with an increasing number of experts. We hypothesize this parameter inefficiency is a result of all experts having equal capacity, which may not adequately meet the varying complexity requirements of different tokens or tasks. In light of this, we propose Stratified Mixture of Experts (SMoE) models, which feature a stratified structure and can assign dynamic capacity to different tokens. We demonstrate the effectiveness of SMoE on three multilingual machine translation benchmarks, containing 4, 15, and 94 language pairs, respectively. We show that SMoE outperforms multiple state-of-the-art MoE models with the same or fewer parameters.

Towards Being Parameter-Efficient: A Stratified Sparsely Activated Transformer with Dynamic Capacity

findings-emnlp.856

2305.02176

https://huggingface.co/papers/2305.02176

Poster

https://aclanthology.org/2023.findings-emnlp.857.bib

https://aclanthology.org/2023.findings-emnlp.857/

@inproceedings{singh-etal-2023-misery, title = "Misery Loves Complexity: Exploring Linguistic Complexity in the Context of Emotion Detection", author = "Singh, Pranaydeep and De Bruyne, Luna and De Clercq, Orph{\'e}e and Lefever, Els", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.857", doi = "10.18653/v1/2023.findings-emnlp.857", pages = "12871--12880", abstract = "Given the omnipresence of social media in our society, thoughts and opinions are being shared online in an unprecedented manner. This means that both positive and negative emotions can be equally and freely expressed. However, the negativity bias posits that human beings are inherently drawn to and more moved by negativity and, as a consequence, negative emotions get more traffic. Correspondingly, when writing about emotions this negativity bias could lead to expressions of negative emotions that are linguistically more complex. In this paper, we attempt to use readability and linguistic complexity metrics to better understand the manifestation of emotions on social media platforms like Reddit based on the widely-used GoEmotions dataset. We demonstrate that according to most metrics, negative emotions indeed tend to generate more complex text than positive emotions. In addition, we examine whether a higher complexity hampers the automatic identification of emotions. To answer this question, we fine-tuned three state-of-the-art transformers (BERT, RoBERTa, and SpanBERT) on the same emotion detection dataset. We demonstrate that these models often fail to predict emotions for the more complex texts. More advanced LLMs like RoBERTa and SpanBERT also fail to improve by significant margins on complex samples. This calls for a more nuanced interpretation of the emotion detection performance of transformer models. We make the automatically annotated data available for further research at: https://huggingface.co/datasets/pranaydeeps/CAMEO", }

Given the omnipresence of social media in our society, thoughts and opinions are being shared online in an unprecedented manner. This means that both positive and negative emotions can be equally and freely expressed. However, the negativity bias posits that human beings are inherently drawn to and more moved by negativity and, as a consequence, negative emotions get more traffic. Correspondingly, when writing about emotions this negativity bias could lead to expressions of negative emotions that are linguistically more complex. In this paper, we attempt to use readability and linguistic complexity metrics to better understand the manifestation of emotions on social media platforms like Reddit based on the widely-used GoEmotions dataset. We demonstrate that according to most metrics, negative emotions indeed tend to generate more complex text than positive emotions. In addition, we examine whether a higher complexity hampers the automatic identification of emotions. To answer this question, we fine-tuned three state-of-the-art transformers (BERT, RoBERTa, and SpanBERT) on the same emotion detection dataset. We demonstrate that these models often fail to predict emotions for the more complex texts. More advanced LLMs like RoBERTa and SpanBERT also fail to improve by significant margins on complex samples. This calls for a more nuanced interpretation of the emotion detection performance of transformer models. We make the automatically annotated data available for further research at: https://huggingface.co/datasets/pranaydeeps/CAMEO

Misery Loves Complexity: Exploring Linguistic Complexity in the Context of Emotion Detection

findings-emnlp.857

Poster

https://aclanthology.org/2023.findings-emnlp.858.bib

https://aclanthology.org/2023.findings-emnlp.858/

@inproceedings{chen-ding-2023-probing, title = "Probing the {``}Creativity{''} of Large Language Models: Can models produce divergent semantic association?", author = "Chen, Honghua and Ding, Nai", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.858", doi = "10.18653/v1/2023.findings-emnlp.858", pages = "12881--12888", abstract = "Large language models possess remarkable capacity for processing language, but it remains unclear whether these models can further generate creative content. The present study aims to investigate the creative thinking of large language models through a cognitive perspective. We utilize the divergent association task (DAT), an objective measurement of creativity that asks models to generate unrelated words and calculates the semantic distance between them. We compare the results across different models and decoding strategies. Our findings indicate that: (1) When using the greedy search strategy, GPT-4 outperforms 96{\%} of humans, while GPT-3.5-turbo exceeds the average human level. (2) Stochastic sampling and temperature scaling are effective to obtain higher DAT scores for models except GPT-4, but face a trade-off between creativity and stability. These results imply that advanced large language models have divergent semantic associations, which is a fundamental process underlying creativity.", }

Large language models possess remarkable capacity for processing language, but it remains unclear whether these models can further generate creative content. The present study aims to investigate the creative thinking of large language models through a cognitive perspective. We utilize the divergent association task (DAT), an objective measurement of creativity that asks models to generate unrelated words and calculates the semantic distance between them. We compare the results across different models and decoding strategies. Our findings indicate that: (1) When using the greedy search strategy, GPT-4 outperforms 96{\%} of humans, while GPT-3.5-turbo exceeds the average human level. (2) Stochastic sampling and temperature scaling are effective to obtain higher DAT scores for models except GPT-4, but face a trade-off between creativity and stability. These results imply that advanced large language models have divergent semantic associations, which is a fundamental process underlying creativity.

Probing the “Creativity” of Large Language Models: Can models produce divergent semantic association?

findings-emnlp.858

Poster

https://aclanthology.org/2023.findings-emnlp.859.bib

https://aclanthology.org/2023.findings-emnlp.859/

@inproceedings{iyer-etal-2023-code, title = "Code-Switching with Word Senses for Pretraining in Neural Machine Translation", author = "Iyer, Vivek and Barba, Edoardo and Birch, Alexandra and Pan, Jeff and Navigli, Roberto", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.859", doi = "10.18653/v1/2023.findings-emnlp.859", pages = "12889--12901", abstract = "Lexical ambiguity is a significant and pervasive challenge in Neural Machine Translation (NMT), with many state-of-the-art (SOTA) NMT systems struggling to handle polysemous words (Campolungo et al., 2022). The same holds for the NMT pretraining paradigm of denoising synthetic {``}code-switched{''} text (Pan et al., 2021; Iyer et al., 2023), where word senses are ignored in the noising stage {--} leading to harmful sense biases in the pretraining data that are subsequently inherited by the resulting models. In this work, we introduce Word Sense Pretraining for Neural Machine Translation (WSP-NMT) - an end-to-end approach for pretraining multilingual NMT models leveraging word sense-specific information from Knowledge Bases. Our experiments show significant improvements in overall translation quality. Then, we show the robustness of our approach to scale to various challenging data and resource-scarce scenarios and, finally, report fine-grained accuracy improvements on the DiBiMT disambiguation benchmark. Our studies yield interesting and novel insights into the merits and challenges of integrating word sense information and structured knowledge in multilingual pretraining for NMT.", }

Lexical ambiguity is a significant and pervasive challenge in Neural Machine Translation (NMT), with many state-of-the-art (SOTA) NMT systems struggling to handle polysemous words (Campolungo et al., 2022). The same holds for the NMT pretraining paradigm of denoising synthetic {``}code-switched{''} text (Pan et al., 2021; Iyer et al., 2023), where word senses are ignored in the noising stage {--} leading to harmful sense biases in the pretraining data that are subsequently inherited by the resulting models. In this work, we introduce Word Sense Pretraining for Neural Machine Translation (WSP-NMT) - an end-to-end approach for pretraining multilingual NMT models leveraging word sense-specific information from Knowledge Bases. Our experiments show significant improvements in overall translation quality. Then, we show the robustness of our approach to scale to various challenging data and resource-scarce scenarios and, finally, report fine-grained accuracy improvements on the DiBiMT disambiguation benchmark. Our studies yield interesting and novel insights into the merits and challenges of integrating word sense information and structured knowledge in multilingual pretraining for NMT.

Code-Switching with Word Senses for Pretraining in Neural Machine Translation

findings-emnlp.859

2310.14050

Poster

https://aclanthology.org/2023.findings-emnlp.860.bib

https://aclanthology.org/2023.findings-emnlp.860/

@inproceedings{huang-etal-2023-diffusionsl, title = "{D}iffusion{SL}: Sequence Labeling via Tag Diffusion Process", author = "Huang, Ziyang and Cao, Pengfei and Zhao, Jun and Liu, Kang", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.860", doi = "10.18653/v1/2023.findings-emnlp.860", pages = "12902--12920", abstract = "Sequence Labeling (SL) is long-standing in Natural Language Processing (NLP). Traditionally, discriminative models have been widely used to capture the conditional distribution of sequence tags, rather than generative models. In this paper, we present DiffusionSL, a framework that utilizes a conditional discrete diffusion model for generating discrete tag data, resulting in a Tag Diffusion Process. We treat the natural language sequence as the conditional signal and the sequence tags as the generation target, iteratively refining the noisy tags to obtain clean ones. To address the discreteness issue, we propose the Bit-Tag Converter (BTConverter) to model the target in continuous data space. Furthermore, we introduce the Bit Diffusion Transformer (BitDiT) to model the process of noise elimination. Leveraging the powerful iterative refinement capability of the diffusion model, DiffusionSL achieves superior performance against previous state-of-the-art (SOTA) baselines and outperforms gpt-3.5-turbo significantly across multiple benchmark datasets and various tasks.", }

Sequence Labeling (SL) is long-standing in Natural Language Processing (NLP). Traditionally, discriminative models have been widely used to capture the conditional distribution of sequence tags, rather than generative models. In this paper, we present DiffusionSL, a framework that utilizes a conditional discrete diffusion model for generating discrete tag data, resulting in a Tag Diffusion Process. We treat the natural language sequence as the conditional signal and the sequence tags as the generation target, iteratively refining the noisy tags to obtain clean ones. To address the discreteness issue, we propose the Bit-Tag Converter (BTConverter) to model the target in continuous data space. Furthermore, we introduce the Bit Diffusion Transformer (BitDiT) to model the process of noise elimination. Leveraging the powerful iterative refinement capability of the diffusion model, DiffusionSL achieves superior performance against previous state-of-the-art (SOTA) baselines and outperforms gpt-3.5-turbo significantly across multiple benchmark datasets and various tasks.

DiffusionSL: Sequence Labeling via Tag Diffusion Process

findings-emnlp.860

Poster

https://aclanthology.org/2023.findings-emnlp.861.bib

https://aclanthology.org/2023.findings-emnlp.861/

@inproceedings{ravi-etal-2023-comet, title = "{COMET}-{M}: Reasoning about Multiple Events in Complex Sentences", author = "Ravi, Sahithya and Ng, Raymond and Shwartz, Vered", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.861", doi = "10.18653/v1/2023.findings-emnlp.861", pages = "12921--12937", abstract = "Understanding the speaker{'}s intended meaning often involves drawing commonsense inferences to reason about what is not stated explicitly. In multi-event sentences, it requires understanding the relationships between events based on contextual knowledge. We propose COMET-M (Multi-Event), an event-centric commonsense model capable of generating commonsense inferences for a target event within a complex sentence. COMET-M builds upon COMET (Bosselut et al., 2019), which excels at generating event-centric inferences for simple sentences, but struggles with the complexity of multi-event sentences prevalent in natural text. To overcome this limitation, we curate a Multi-Event Inference (MEI) dataset of 35K human-written inferences. We train COMET-M on the human-written inferences and also create baselines using automatically labeled examples. Experimental results demonstrate the significant performance improvement of COMET-M over COMET in generating multi-event inferences. Moreover, COMET-M successfully produces distinct inferences for each target event, taking the complete context into consideration. COMET-M holds promise for downstream tasks involving natural text such as coreference resolution, dialogue, and story understanding.", }

Understanding the speaker{'}s intended meaning often involves drawing commonsense inferences to reason about what is not stated explicitly. In multi-event sentences, it requires understanding the relationships between events based on contextual knowledge. We propose COMET-M (Multi-Event), an event-centric commonsense model capable of generating commonsense inferences for a target event within a complex sentence. COMET-M builds upon COMET (Bosselut et al., 2019), which excels at generating event-centric inferences for simple sentences, but struggles with the complexity of multi-event sentences prevalent in natural text. To overcome this limitation, we curate a Multi-Event Inference (MEI) dataset of 35K human-written inferences. We train COMET-M on the human-written inferences and also create baselines using automatically labeled examples. Experimental results demonstrate the significant performance improvement of COMET-M over COMET in generating multi-event inferences. Moreover, COMET-M successfully produces distinct inferences for each target event, taking the complete context into consideration. COMET-M holds promise for downstream tasks involving natural text such as coreference resolution, dialogue, and story understanding.

COMET-M: Reasoning about Multiple Events in Complex Sentences

findings-emnlp.861

2305.14617

Poster

https://aclanthology.org/2023.findings-emnlp.862.bib

https://aclanthology.org/2023.findings-emnlp.862/

@inproceedings{gantt-etal-2023-event, title = "On Event Individuation for Document-Level Information Extraction", author = "Gantt, William and Kriz, Reno and Chen, Yunmo and Vashishtha, Siddharth and White, Aaron", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.862", doi = "10.18653/v1/2023.findings-emnlp.862", pages = "12938--12958", abstract = "As information extraction (IE) systems have grown more adept at processing whole documents, the classic task of *template filling* has seen renewed interest as a benchmark for document-level IE. In this position paper, we call into question the suitability of template filling for this purpose. We argue that the task demands definitive answers to thorny questions of *event individuation* {---} the problem of distinguishing distinct events {---} about which even human experts disagree. Through an annotation study and error analysis, we show that this raises concerns about the usefulness of template filling metrics, the quality of datasets for the task, and the ability of models to learn it. Finally, we consider possible solutions.", }

As information extraction (IE) systems have grown more adept at processing whole documents, the classic task of *template filling* has seen renewed interest as a benchmark for document-level IE. In this position paper, we call into question the suitability of template filling for this purpose. We argue that the task demands definitive answers to thorny questions of *event individuation* {---} the problem of distinguishing distinct events {---} about which even human experts disagree. Through an annotation study and error analysis, we show that this raises concerns about the usefulness of template filling metrics, the quality of datasets for the task, and the ability of models to learn it. Finally, we consider possible solutions.

On Event Individuation for Document-Level Information Extraction

findings-emnlp.862

2212.09702

Poster

https://aclanthology.org/2023.findings-emnlp.863.bib

https://aclanthology.org/2023.findings-emnlp.863/

@inproceedings{kim-etal-2023-aniee, title = "{A}ni{EE}: A Dataset of Animal Experimental Literature for Event Extraction", author = "Kim, Dohee and Yoo, Ra and Yang, Soyoung and Yang, Hee and Choo, Jaegul", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.863", doi = "10.18653/v1/2023.findings-emnlp.863", pages = "12959--12971", abstract = "Event extraction (EE), as a crucial information extraction (IE) task, aims to identify event triggers and their associated arguments from unstructured text, subsequently classifying them into pre-defined types and roles. In the biomedical domain, EE is widely used to extract complex structures representing biological events from literature. Due to the complicated semantics and specialized domain knowledge, it is challenging to construct biomedical event extraction datasets. Additionally, most existing biomedical EE datasets primarily focus on cell experiments or the overall experimental procedures. Therefore, we introduce AniEE, an event extraction dataset concentrated on the animal experiment stage. We establish a novel animal experiment customized entity and event scheme in collaboration with domain experts. We then create an expert-annotated high-quality dataset containing discontinuous entities and nested events and evaluate our dataset on the recent outstanding NER and EE models.", }

Event extraction (EE), as a crucial information extraction (IE) task, aims to identify event triggers and their associated arguments from unstructured text, subsequently classifying them into pre-defined types and roles. In the biomedical domain, EE is widely used to extract complex structures representing biological events from literature. Due to the complicated semantics and specialized domain knowledge, it is challenging to construct biomedical event extraction datasets. Additionally, most existing biomedical EE datasets primarily focus on cell experiments or the overall experimental procedures. Therefore, we introduce AniEE, an event extraction dataset concentrated on the animal experiment stage. We establish a novel animal experiment customized entity and event scheme in collaboration with domain experts. We then create an expert-annotated high-quality dataset containing discontinuous entities and nested events and evaluate our dataset on the recent outstanding NER and EE models.

AniEE: A Dataset of Animal Experimental Literature for Event Extraction

findings-emnlp.863

Poster

https://aclanthology.org/2023.findings-emnlp.864.bib

https://aclanthology.org/2023.findings-emnlp.864/

@inproceedings{jansen-2023-words, title = "From Words to Wires: Generating Functioning Electronic Devices from Natural Language Descriptions", author = "Jansen, Peter", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.864", doi = "10.18653/v1/2023.findings-emnlp.864", pages = "12972--12990", abstract = "In this work, we show that contemporary language models have a previously unknown skill {--} the capacity for electronic circuit design from high-level textual descriptions, akin to code generation. We introduce two benchmarks: PINS100, assessing model knowledge of electrical components, and MICRO25, evaluating a model{'}s capability to design common microcontroller circuits and code in the Arduino ecosystem that involve input, output, sensors, motors, protocols, and logic {--} with models such as GPT-4 and Claude-V1 achieving between 60{\%} to 96{\%} Pass@1 on generating full devices. We include six case studies of using language models as a design assistant for moderately complex devices, such as a radiation-powered random number generator, an emoji keyboard, a visible spectrometer, and several assistive devices, while offering a qualitative analysis performance, outlining evaluation challenges, and suggesting areas of development to improve complex circuit design and practical utility. With this work, we aim to spur research at the juncture of natural language processing and electronic design.", }

In this work, we show that contemporary language models have a previously unknown skill {--} the capacity for electronic circuit design from high-level textual descriptions, akin to code generation. We introduce two benchmarks: PINS100, assessing model knowledge of electrical components, and MICRO25, evaluating a model{'}s capability to design common microcontroller circuits and code in the Arduino ecosystem that involve input, output, sensors, motors, protocols, and logic {--} with models such as GPT-4 and Claude-V1 achieving between 60{\%} to 96{\%} Pass@1 on generating full devices. We include six case studies of using language models as a design assistant for moderately complex devices, such as a radiation-powered random number generator, an emoji keyboard, a visible spectrometer, and several assistive devices, while offering a qualitative analysis performance, outlining evaluation challenges, and suggesting areas of development to improve complex circuit design and practical utility. With this work, we aim to spur research at the juncture of natural language processing and electronic design.

From Words to Wires: Generating Functioning Electronic Devices from Natural Language Descriptions

findings-emnlp.864

2305.14874

Poster

https://aclanthology.org/2023.findings-emnlp.865.bib

https://aclanthology.org/2023.findings-emnlp.865/

@inproceedings{zhang-etal-2023-data, title = "Data-efficient Active Learning for Structured Prediction with Partial Annotation and Self-Training", author = "Zhang, Zhisong and Strubell, Emma and Hovy, Eduard", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.865", doi = "10.18653/v1/2023.findings-emnlp.865", pages = "12991--13008", abstract = "In this work we propose a pragmatic method that reduces the annotation cost for structured label spaces using active learning. Our approach leverages partial annotation, which reduces labeling costs for structured outputs by selecting only the most informative sub-structures for annotation. We also utilize self-training to incorporate the current model{'}s automatic predictions as pseudo-labels for un-annotated sub-structures. A key challenge in effectively combining partial annotation with self-training to reduce annotation cost is determining which sub-structures to select to label. To address this challenge, we adopt an error estimator to adaptively decide the partial selection ratio according to the current model{'}s capability. In evaluations spanning four structured prediction tasks, we show that our combination of partial annotation and self-training using an adaptive selection ratio reduces annotation cost over strong full annotation baselines under a fair comparison scheme that takes reading time into consideration.", }

In this work we propose a pragmatic method that reduces the annotation cost for structured label spaces using active learning. Our approach leverages partial annotation, which reduces labeling costs for structured outputs by selecting only the most informative sub-structures for annotation. We also utilize self-training to incorporate the current model{'}s automatic predictions as pseudo-labels for un-annotated sub-structures. A key challenge in effectively combining partial annotation with self-training to reduce annotation cost is determining which sub-structures to select to label. To address this challenge, we adopt an error estimator to adaptively decide the partial selection ratio according to the current model{'}s capability. In evaluations spanning four structured prediction tasks, we show that our combination of partial annotation and self-training using an adaptive selection ratio reduces annotation cost over strong full annotation baselines under a fair comparison scheme that takes reading time into consideration.

Data-efficient Active Learning for Structured Prediction with Partial Annotation and Self-Training

findings-emnlp.865

2305.12634

Poster

https://aclanthology.org/2023.findings-emnlp.866.bib

https://aclanthology.org/2023.findings-emnlp.866/

@inproceedings{luo-etal-2023-explicit, title = "Explicit Alignment and Many-to-many Entailment Based Reasoning for Conversational Machine Reading", author = "Luo, Yangyang and Tian, Shiyu and Yuan, Caixia and Wang, Xiaojie", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.866", doi = "10.18653/v1/2023.findings-emnlp.866", pages = "13009--13022", abstract = "Conversational Machine Reading (CMR) requires answering a user{'}s initial question through multi-turn dialogue interactions based on a given document. Although there exist many effective methods, they largely neglected the alignment between the $\textit{document}$ and the $\textit{user-provided information}$, which significantly affects the intermediate decision-making and subsequent follow-up question generation. To address this issue, we propose a pipeline framework that (1) aligns the aforementioned two sides in an explicit way, (2) makes decisions using a lightweight many-to-many entailment reasoning module, and (3) directly generates follow-up questions based on the document and previously asked questions. Our proposed method achieves state-of-the-art in micro-accuracy and ranks the first place on the public leaderboard of the CMR benchmark dataset ShARC.", }

Conversational Machine Reading (CMR) requires answering a user{'}s initial question through multi-turn dialogue interactions based on a given document. Although there exist many effective methods, they largely neglected the alignment between the $\textit{document}$ and the $\textit{user-provided information}$, which significantly affects the intermediate decision-making and subsequent follow-up question generation. To address this issue, we propose a pipeline framework that (1) aligns the aforementioned two sides in an explicit way, (2) makes decisions using a lightweight many-to-many entailment reasoning module, and (3) directly generates follow-up questions based on the document and previously asked questions. Our proposed method achieves state-of-the-art in micro-accuracy and ranks the first place on the public leaderboard of the CMR benchmark dataset ShARC.

Explicit Alignment and Many-to-many Entailment Based Reasoning for Conversational Machine Reading

findings-emnlp.866

2310.13409

Poster

https://aclanthology.org/2023.findings-emnlp.867.bib

https://aclanthology.org/2023.findings-emnlp.867/

@inproceedings{ince-etal-2023-harnessing, title = "Harnessing Dataset Cartography for Improved Compositional Generalization in Transformers", author = "{\.I}nce, Osman and Zeraati, Tanin and Yagcioglu, Semih and Yaghoobzadeh, Yadollah and Erdem, Erkut and Erdem, Aykut", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.867", doi = "10.18653/v1/2023.findings-emnlp.867", pages = "13023--13041", abstract = "Neural networks have revolutionized language modeling and excelled in various downstream tasks. However, the extent to which these models achieve compositional generalization comparable to human cognitive abilities remains a topic of debate. While existing approaches in the field have mainly focused on novel architectures and alternative learning paradigms, we introduce a pioneering method harnessing the power of dataset cartography (Swayamdipta et al., 2020). By strategically identifying a subset of compositional generalization data using this approach, we achieve a remarkable improvement in model accuracy, yielding enhancements of up to 10{\%} on CFQ and COGS datasets. Notably, our technique incorporates dataset cartography as a curriculum learning criterion, eliminating the need for hyperparameter tuning while consistently achieving superior performance. Our findings highlight the untapped potential of dataset cartography in unleashing the full capabilities of compositional generalization within Transformer models.", }

Neural networks have revolutionized language modeling and excelled in various downstream tasks. However, the extent to which these models achieve compositional generalization comparable to human cognitive abilities remains a topic of debate. While existing approaches in the field have mainly focused on novel architectures and alternative learning paradigms, we introduce a pioneering method harnessing the power of dataset cartography (Swayamdipta et al., 2020). By strategically identifying a subset of compositional generalization data using this approach, we achieve a remarkable improvement in model accuracy, yielding enhancements of up to 10{\%} on CFQ and COGS datasets. Notably, our technique incorporates dataset cartography as a curriculum learning criterion, eliminating the need for hyperparameter tuning while consistently achieving superior performance. Our findings highlight the untapped potential of dataset cartography in unleashing the full capabilities of compositional generalization within Transformer models.

Harnessing Dataset Cartography for Improved Compositional Generalization in Transformers

findings-emnlp.867

2310.12118

Poster

https://aclanthology.org/2023.findings-emnlp.868.bib

https://aclanthology.org/2023.findings-emnlp.868/

@inproceedings{gao-etal-2023-roles, title = "Roles of Scaling and Instruction Tuning in Language Perception: Model vs. Human Attention", author = "Gao, Changjiang and Huang, Shujian and Li, Jixing and Chen, Jiajun", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.868", doi = "10.18653/v1/2023.findings-emnlp.868", pages = "13042--13055", abstract = "Recent large language models (LLMs) have revealed strong abilities to understand natural language. Since most of them share the same basic structure, i.e. the transformer block, possible contributors to their success in the training process are scaling and instruction tuning. However, how these factors affect the models{'} language perception is unclear. This work compares the self-attention of several existing LLMs (LLaMA, Alpaca and Vicuna) in different sizes (7B, 13B, 30B, 65B), together with eye saccade, an aspect of human reading attention, to assess the effect of scaling and instruction tuning on language perception. Results show that scaling enhances the human resemblance and improves the effective attention by reducing the trivial pattern reliance, while instruction tuning does not. However, instruction tuning significantly enhances the models{'} sensitivity to instructions. We also find that current LLMs are consistently closer to non-native than native speakers in attention, suggesting a sub-optimal language perception of all models. Our code and data used in the analysis is available on GitHub.", }

Recent large language models (LLMs) have revealed strong abilities to understand natural language. Since most of them share the same basic structure, i.e. the transformer block, possible contributors to their success in the training process are scaling and instruction tuning. However, how these factors affect the models{'} language perception is unclear. This work compares the self-attention of several existing LLMs (LLaMA, Alpaca and Vicuna) in different sizes (7B, 13B, 30B, 65B), together with eye saccade, an aspect of human reading attention, to assess the effect of scaling and instruction tuning on language perception. Results show that scaling enhances the human resemblance and improves the effective attention by reducing the trivial pattern reliance, while instruction tuning does not. However, instruction tuning significantly enhances the models{'} sensitivity to instructions. We also find that current LLMs are consistently closer to non-native than native speakers in attention, suggesting a sub-optimal language perception of all models. Our code and data used in the analysis is available on GitHub.

Roles of Scaling and Instruction Tuning in Language Perception: Model vs. Human Attention

findings-emnlp.868

2310.19084

https://huggingface.co/papers/2310.19084

Poster

https://aclanthology.org/2023.findings-emnlp.869.bib

https://aclanthology.org/2023.findings-emnlp.869/

@inproceedings{fan-he-2023-efficient, title = "Efficient Data Learning for Open Information Extraction with Pre-trained Language Models", author = "Fan, Zhiyuan and He, Shizhu", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.869", doi = "10.18653/v1/2023.findings-emnlp.869", pages = "13056--13063", abstract = "Open Information Extraction (OpenIE) is a fundamental yet challenging task in Natural Language Processing, which involves extracting all triples (subject, predicate, object) from a given sentence. While labelling-based methods have their merits, generation-based techniques offer unique advantages, such as the ability to generate tokens not present in the original sentence. However, these generation-based methods often require a significant amount of training data to learn the task form of OpenIE and substantial training time to overcome slow model convergence due to the order penalty. In this paper, we introduce a novel framework, OK-IE, that ingeniously transforms the task form of OpenIE into the pre-training task form of the T5 model, thereby reducing the need for extensive training data. Furthermore, we introduce an innovative concept of {`}anchors{'} to control the sequence of model outputs, effectively eliminating the impact of order penalty on model convergence and significantly reducing training time. Experimental results indicate that, compared to previous SOTA methods, OK-IE requires only 1/100 of the training data (900 instances) and 1/120 of the training time (3 minutes) to achieve comparable results.", }

Open Information Extraction (OpenIE) is a fundamental yet challenging task in Natural Language Processing, which involves extracting all triples (subject, predicate, object) from a given sentence. While labelling-based methods have their merits, generation-based techniques offer unique advantages, such as the ability to generate tokens not present in the original sentence. However, these generation-based methods often require a significant amount of training data to learn the task form of OpenIE and substantial training time to overcome slow model convergence due to the order penalty. In this paper, we introduce a novel framework, OK-IE, that ingeniously transforms the task form of OpenIE into the pre-training task form of the T5 model, thereby reducing the need for extensive training data. Furthermore, we introduce an innovative concept of {`}anchors{'} to control the sequence of model outputs, effectively eliminating the impact of order penalty on model convergence and significantly reducing training time. Experimental results indicate that, compared to previous SOTA methods, OK-IE requires only 1/100 of the training data (900 instances) and 1/120 of the training time (3 minutes) to achieve comparable results.

Efficient Data Learning for Open Information Extraction with Pre-trained Language Models

findings-emnlp.869

2310.15021

Poster

https://aclanthology.org/2023.findings-emnlp.870.bib

https://aclanthology.org/2023.findings-emnlp.870/

@inproceedings{zhou-etal-2023-survival, title = "Survival of the Most Influential Prompts: Efficient Black-Box Prompt Search via Clustering and Pruning", author = "Zhou, Han and Wan, Xingchen and Vuli{\'c}, Ivan and Korhonen, Anna", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.870", doi = "10.18653/v1/2023.findings-emnlp.870", pages = "13064--13077", abstract = "Prompt-based learning has been an effective paradigm for large pretrained language models (LLM), enabling few-shot or even zero-shot learning. Black-box prompt search has received growing interest recently for its distinctive properties of gradient-free optimization, proven particularly useful and powerful for model-as-a-service usage. However, the discrete nature and the complexity of combinatorial optimization hinder the efficiency of modern black-box approaches. Despite extensive research on search algorithms, the crucial aspect of search space design and optimization has been largely overlooked. In this paper, we first conduct a sensitivity analysis by prompting LLM, revealing that only a small number of tokens exert a disproportionate amount of influence on LLM predictions. Leveraging this insight, we propose the Clustering and Pruning for Efficient Black-box Prompt Search (ClaPS), a simple black-box search method that first clusters and prunes the search space to focus exclusively on influential prompt tokens. By employing even simple search methods within the pruned search space, ClaPS achieves state-of-the-art performance across various tasks and LLMs, surpassing the performance of complex approaches while significantly reducing search costs. Our findings underscore the critical role of search space design and optimization in enhancing both the usefulness and the efficiency of black-box prompt-based learning.", }

Prompt-based learning has been an effective paradigm for large pretrained language models (LLM), enabling few-shot or even zero-shot learning. Black-box prompt search has received growing interest recently for its distinctive properties of gradient-free optimization, proven particularly useful and powerful for model-as-a-service usage. However, the discrete nature and the complexity of combinatorial optimization hinder the efficiency of modern black-box approaches. Despite extensive research on search algorithms, the crucial aspect of search space design and optimization has been largely overlooked. In this paper, we first conduct a sensitivity analysis by prompting LLM, revealing that only a small number of tokens exert a disproportionate amount of influence on LLM predictions. Leveraging this insight, we propose the Clustering and Pruning for Efficient Black-box Prompt Search (ClaPS), a simple black-box search method that first clusters and prunes the search space to focus exclusively on influential prompt tokens. By employing even simple search methods within the pruned search space, ClaPS achieves state-of-the-art performance across various tasks and LLMs, surpassing the performance of complex approaches while significantly reducing search costs. Our findings underscore the critical role of search space design and optimization in enhancing both the usefulness and the efficiency of black-box prompt-based learning.

Survival of the Most Influential Prompts: Efficient Black-Box Prompt Search via Clustering and Pruning

findings-emnlp.870

2310.12774

https://huggingface.co/papers/2310.12774

Poster

https://aclanthology.org/2023.findings-emnlp.871.bib

https://aclanthology.org/2023.findings-emnlp.871/

@inproceedings{yang-etal-2023-towards-zero, title = "Towards Zero-shot Learning for End-to-end Cross-modal Translation Models", author = "Yang, Jichen and Fan, Kai and Liao, Minpeng and Chen, Boxing and Huang, Zhongqiang", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.871", doi = "10.18653/v1/2023.findings-emnlp.871", pages = "13078--13087", abstract = "One of the main problems in speech translation is the mismatches between different modalities. The second problem, scarcity of parallel data covering multiple modalities, means that the end-to-end multi-modal models tend to perform worse than cascade models, although there are exceptions under favorable conditions. To address these problems, we propose an end-to-end zero-shot speech translation model, connecting two pre-trained uni-modality modules via word rotator{'}s distance. The model retains the ability of zero-shot, which is like cascade models, and also can be trained in an end-to-end style to avoid error propagation. Our comprehensive experiments on the MuST-C benchmarks show that our end-to-end zero-shot approach performs better than or as well as those of the CTC-based cascade models and that our end-to-end model with supervised training also matches the latest baselines.", }

One of the main problems in speech translation is the mismatches between different modalities. The second problem, scarcity of parallel data covering multiple modalities, means that the end-to-end multi-modal models tend to perform worse than cascade models, although there are exceptions under favorable conditions. To address these problems, we propose an end-to-end zero-shot speech translation model, connecting two pre-trained uni-modality modules via word rotator{'}s distance. The model retains the ability of zero-shot, which is like cascade models, and also can be trained in an end-to-end style to avoid error propagation. Our comprehensive experiments on the MuST-C benchmarks show that our end-to-end zero-shot approach performs better than or as well as those of the CTC-based cascade models and that our end-to-end model with supervised training also matches the latest baselines.

Towards Zero-shot Learning for End-to-end Cross-modal Translation Models

findings-emnlp.871

Poster

https://aclanthology.org/2023.findings-emnlp.872.bib

https://aclanthology.org/2023.findings-emnlp.872/

@inproceedings{zhang-etal-2023-llmaaa, title = "{LLM}a{AA}: Making Large Language Models as Active Annotators", author = "Zhang, Ruoyu and Li, Yanzeng and Ma, Yongliang and Zhou, Ming and Zou, Lei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.872", doi = "10.18653/v1/2023.findings-emnlp.872", pages = "13088--13103", abstract = "Prevalent supervised learning methods in natural language processing (NLP) are notoriously data-hungry, which demand large amounts of high-quality annotated data. In practice, acquiring such data is a costly endeavor. Recently, the superior few-shot performance of large language models (LLMs) has propelled the development of dataset generation, where the training data are solely synthesized from LLMs. However, such an approach usually suffers from low-quality issues, and requires orders of magnitude more labeled data to achieve satisfactory performance. To fully exploit the potential of LLMs and make use of massive unlabeled data, we propose LLMaAA, which takes LLMs as annotators and puts them into an active learning loop to determine what to annotate efficiently. To learn robustly with pseudo labels, we optimize both the annotation and training processes: (1) we draw $k$-NN examples from a small demonstration pool as in-context examples, and (2) we adopt the example reweighting technique to assign training samples with learnable weights. Compared with previous approaches, LLMaAA features both efficiency and reliability. We conduct experiments and analysis on two classic NLP tasks, named entity recognition and relation extraction. With LLMaAA, task-specific models trained from LLM-generated labels can outperform the teacher within only hundreds of annotated examples, which is much more cost-effective than other baselines.", }

Prevalent supervised learning methods in natural language processing (NLP) are notoriously data-hungry, which demand large amounts of high-quality annotated data. In practice, acquiring such data is a costly endeavor. Recently, the superior few-shot performance of large language models (LLMs) has propelled the development of dataset generation, where the training data are solely synthesized from LLMs. However, such an approach usually suffers from low-quality issues, and requires orders of magnitude more labeled data to achieve satisfactory performance. To fully exploit the potential of LLMs and make use of massive unlabeled data, we propose LLMaAA, which takes LLMs as annotators and puts them into an active learning loop to determine what to annotate efficiently. To learn robustly with pseudo labels, we optimize both the annotation and training processes: (1) we draw $k$-NN examples from a small demonstration pool as in-context examples, and (2) we adopt the example reweighting technique to assign training samples with learnable weights. Compared with previous approaches, LLMaAA features both efficiency and reliability. We conduct experiments and analysis on two classic NLP tasks, named entity recognition and relation extraction. With LLMaAA, task-specific models trained from LLM-generated labels can outperform the teacher within only hundreds of annotated examples, which is much more cost-effective than other baselines.

LLMaAA: Making Large Language Models as Active Annotators

findings-emnlp.872

2310.19596

Poster

https://aclanthology.org/2023.findings-emnlp.873.bib

https://aclanthology.org/2023.findings-emnlp.873/

@inproceedings{singh-etal-2023-nlms, title = "{NLM}s: Augmenting Negation in Language Models", author = "Singh, Rituraj and Kumar, Rahul and Sridhar, Vivek", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.873", doi = "10.18653/v1/2023.findings-emnlp.873", pages = "13104--13116", abstract = "Negation is the fundamental component in a natural language that reverses the semantic meaning of a sentence. It plays an extremely important role across a wide range of applications, yet they are underrepresented in pre-trained language models (LMs), resulting often in wrong inferences. In this work, we try to improve the underlying understanding of the negation in the pre-trained LMs. To augment negation understanding, we propose a language model objective with a weighted cross-entropy loss and elastic weight consolidation regularization. We reduce the mean top 1 error rate for BERT-base to 1.1{\%}, BERT-large to 0.78{\%}, RoBERTA-base to 3.74{\%}, RoBERTA-large to 0.01{\%} on the negated LAMA dataset. It minimizes the BERT error rate by a margin of 8{\%} and also outperform the existing negation models. We also provide empirical evidences that negated augmented models outperform the classical models on original as well as negation benchmarks on natural language inference tasks.", }

Negation is the fundamental component in a natural language that reverses the semantic meaning of a sentence. It plays an extremely important role across a wide range of applications, yet they are underrepresented in pre-trained language models (LMs), resulting often in wrong inferences. In this work, we try to improve the underlying understanding of the negation in the pre-trained LMs. To augment negation understanding, we propose a language model objective with a weighted cross-entropy loss and elastic weight consolidation regularization. We reduce the mean top 1 error rate for BERT-base to 1.1{\%}, BERT-large to 0.78{\%}, RoBERTA-base to 3.74{\%}, RoBERTA-large to 0.01{\%} on the negated LAMA dataset. It minimizes the BERT error rate by a margin of 8{\%} and also outperform the existing negation models. We also provide empirical evidences that negated augmented models outperform the classical models on original as well as negation benchmarks on natural language inference tasks.

NLMs: Augmenting Negation in Language Models

findings-emnlp.873

Poster

https://aclanthology.org/2023.findings-emnlp.874.bib

https://aclanthology.org/2023.findings-emnlp.874/

@inproceedings{tam-etal-2023-parameter, title = "Parameter-Efficient Prompt Tuning Makes Generalized and Calibrated Neural Text Retrievers", author = "Tam, Weng and Liu, Xiao and Ji, Kaixuan and Xue, Lilong and Liu, Jiahua and Li, Tao and Dong, Yuxiao and Tang, Jie", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.874", doi = "10.18653/v1/2023.findings-emnlp.874", pages = "13117--13130", abstract = "Prompt tuning attempts to update few task-specific parameters in pre-trained models. It has achieved comparable performance to fine-tuning of the full parameter set on both language understanding and generation tasks. In this work, we study the problem of prompt tuning for neural text retrievers. We introduce parameter-efficient prompt tuning for text retrieval across in-domain, cross-domain, and cross-topic settings. Through an extensive analysis, we show that the strategy can mitigate the two issues{---}parameter-inefficiency and weak generalizability{---}faced by fine-tuning based retrieval methods. Notably, it can significantly improve the out-of-domain zero-shot generalization of the retrieval models. By updating only 0.1{\%} of the model parameters, the prompt tuning strategy can help retrieval models achieve better generalization performance than traditional methods in which all parameters are updated. Finally, to facilitate research on retrievers{'} cross-topic generalizability, we curate and release an academic retrieval dataset with 18K query-results pairs in 87 topics, making it the largest topic-specific one to date.", }

Prompt tuning attempts to update few task-specific parameters in pre-trained models. It has achieved comparable performance to fine-tuning of the full parameter set on both language understanding and generation tasks. In this work, we study the problem of prompt tuning for neural text retrievers. We introduce parameter-efficient prompt tuning for text retrieval across in-domain, cross-domain, and cross-topic settings. Through an extensive analysis, we show that the strategy can mitigate the two issues{---}parameter-inefficiency and weak generalizability{---}faced by fine-tuning based retrieval methods. Notably, it can significantly improve the out-of-domain zero-shot generalization of the retrieval models. By updating only 0.1{\%} of the model parameters, the prompt tuning strategy can help retrieval models achieve better generalization performance than traditional methods in which all parameters are updated. Finally, to facilitate research on retrievers{'} cross-topic generalizability, we curate and release an academic retrieval dataset with 18K query-results pairs in 87 topics, making it the largest topic-specific one to date.

Parameter-Efficient Prompt Tuning Makes Generalized and Calibrated Neural Text Retrievers

findings-emnlp.874

2207.07087

Poster

https://aclanthology.org/2023.findings-emnlp.875.bib

https://aclanthology.org/2023.findings-emnlp.875/

@inproceedings{yun-etal-2023-x, title = "{X}-{SNS}: Cross-Lingual Transfer Prediction through Sub-Network Similarity", author = "Yun, Taejun and Kim, Jinhyeon and Kang, Deokyeong and Lim, Seonghoon and Kim, Jihoon and Kim, Taeuk", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.875", doi = "10.18653/v1/2023.findings-emnlp.875", pages = "13131--13144", abstract = "Cross-lingual transfer (XLT) is an emergent ability of multilingual language models that preserves their performance on a task to a significant extent when evaluated in languages that were not included in the fine-tuning process. While English, due to its widespread usage, is typically regarded as the primary language for model adaption in various tasks, recent studies have revealed that the efficacy of XLT can be amplified by selecting the most appropriate source languages based on specific conditions. In this work, we propose the utilization of sub-network similarity between two languages as a proxy for predicting the compatibility of the languages in the context of XLT. Our approach is model-oriented, better reflecting the inner workings of foundation models. In addition, it requires only a moderate amount of raw text from candidate languages, distinguishing it from the majority of previous methods that rely on external resources. In experiments, we demonstrate that our method is more effective than baselines across diverse tasks. Specifically, it shows proficiency in ranking candidates for zero-shot XLT, achieving an improvement of 4.6{\%} on average in terms of NDCG@3. We also provide extensive analyses that confirm the utility of sub-networks for XLT prediction.", }

Cross-lingual transfer (XLT) is an emergent ability of multilingual language models that preserves their performance on a task to a significant extent when evaluated in languages that were not included in the fine-tuning process. While English, due to its widespread usage, is typically regarded as the primary language for model adaption in various tasks, recent studies have revealed that the efficacy of XLT can be amplified by selecting the most appropriate source languages based on specific conditions. In this work, we propose the utilization of sub-network similarity between two languages as a proxy for predicting the compatibility of the languages in the context of XLT. Our approach is model-oriented, better reflecting the inner workings of foundation models. In addition, it requires only a moderate amount of raw text from candidate languages, distinguishing it from the majority of previous methods that rely on external resources. In experiments, we demonstrate that our method is more effective than baselines across diverse tasks. Specifically, it shows proficiency in ranking candidates for zero-shot XLT, achieving an improvement of 4.6{\%} on average in terms of NDCG@3. We also provide extensive analyses that confirm the utility of sub-networks for XLT prediction.

X-SNS: Cross-Lingual Transfer Prediction through Sub-Network Similarity

findings-emnlp.875

2310.17166

Poster

https://aclanthology.org/2023.findings-emnlp.876.bib

https://aclanthology.org/2023.findings-emnlp.876/

@inproceedings{sun-etal-2023-noise, title = "Noise-Robust Semi-Supervised Learning for Distantly Supervised Relation Extraction", author = "Sun, Xin and Liu, Qiang and Wu, Shu and Wang, Zilei and Wang, Liang", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.876", doi = "10.18653/v1/2023.findings-emnlp.876", pages = "13145--13157", abstract = "Distantly supervised relation extraction (DSRE) aims to extract relational facts from texts but suffers from noisy instances. To mitigate the influence of noisy labels, current methods typically use the Multi-Instance-Learning framework to extract relations for each bag. However, these approaches are not capable of extracting relation labels for individual sentences. Several studies have focused on sentence-level DSRE to solve the above problem. These studies primarily aim to develop methods for identifying noisy samples and filtering them out to mitigate the impact of noise. However, discarding noisy samples directly leads to the loss of useful information. To this end, we propose SSLRE, a novel Semi-Supervised-Learning Relation Extraction framework for sentence-level DSRE. We discard only the labels of the noisy samples and utilize these instances without labels as unlabeled samples. Our SSLRE framework utilizes a weighted K-NN graph to select confident samples as labeled data and the rest as unlabeled. We then design a robust semi-supervised learning framework that can efficiently handle remaining label noise present in the labeled dataset, while also making effective use of unlabeled samples. Based on our experiments on two real-world datasets, the SSLRE framework we proposed has achieved significant enhancements in sentence-level relation extraction performance compared to the existing state-of-the-art methods. Moreover, it has also attained a state-of-the-art level of performance in bag-level relation extraction with ONE aggregation strategy.", }

Distantly supervised relation extraction (DSRE) aims to extract relational facts from texts but suffers from noisy instances. To mitigate the influence of noisy labels, current methods typically use the Multi-Instance-Learning framework to extract relations for each bag. However, these approaches are not capable of extracting relation labels for individual sentences. Several studies have focused on sentence-level DSRE to solve the above problem. These studies primarily aim to develop methods for identifying noisy samples and filtering them out to mitigate the impact of noise. However, discarding noisy samples directly leads to the loss of useful information. To this end, we propose SSLRE, a novel Semi-Supervised-Learning Relation Extraction framework for sentence-level DSRE. We discard only the labels of the noisy samples and utilize these instances without labels as unlabeled samples. Our SSLRE framework utilizes a weighted K-NN graph to select confident samples as labeled data and the rest as unlabeled. We then design a robust semi-supervised learning framework that can efficiently handle remaining label noise present in the labeled dataset, while also making effective use of unlabeled samples. Based on our experiments on two real-world datasets, the SSLRE framework we proposed has achieved significant enhancements in sentence-level relation extraction performance compared to the existing state-of-the-art methods. Moreover, it has also attained a state-of-the-art level of performance in bag-level relation extraction with ONE aggregation strategy.

Noise-Robust Semi-Supervised Learning for Distantly Supervised Relation Extraction

findings-emnlp.876

Poster

https://aclanthology.org/2023.findings-emnlp.877.bib

https://aclanthology.org/2023.findings-emnlp.877/

@inproceedings{shani-etal-2023-towards, title = "Towards Concept-Aware Large Language Models", author = "Shani, Chen and Vreeken, Jilles and Shahaf, Dafna", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.877", doi = "10.18653/v1/2023.findings-emnlp.877", pages = "13158--13170", abstract = "Concepts play a pivotal role in various human cognitive functions, including learning, reasoning and communication. However, there is very little work on endowing machines with the ability to form and reason with concepts. In particular, state-of-the-art large language models (LLMs) work at the level of tokens, not concepts. In this work, we analyze how well contemporary LLMs capture human concepts and their structure. We then discuss ways to develop concept-aware LLMs, taking place at different stages of the pipeline. We sketch a method for pretraining LLMs using concepts, and also explore the simpler approach that uses the output of existing LLMs. Despite its simplicity, our proof-of-concept is shown to better match human intuition, as well as improve the robustness of predictions. These preliminary results underscore the promise of concept-aware LLMs.", }

Concepts play a pivotal role in various human cognitive functions, including learning, reasoning and communication. However, there is very little work on endowing machines with the ability to form and reason with concepts. In particular, state-of-the-art large language models (LLMs) work at the level of tokens, not concepts. In this work, we analyze how well contemporary LLMs capture human concepts and their structure. We then discuss ways to develop concept-aware LLMs, taking place at different stages of the pipeline. We sketch a method for pretraining LLMs using concepts, and also explore the simpler approach that uses the output of existing LLMs. Despite its simplicity, our proof-of-concept is shown to better match human intuition, as well as improve the robustness of predictions. These preliminary results underscore the promise of concept-aware LLMs.

Towards Concept-Aware Large Language Models

findings-emnlp.877

2311.01866

Poster

https://aclanthology.org/2023.findings-emnlp.878.bib

https://aclanthology.org/2023.findings-emnlp.878/

@inproceedings{lai-etal-2023-chatgpt, title = "{C}hat{GPT} Beyond {E}nglish: Towards a Comprehensive Evaluation of Large Language Models in Multilingual Learning", author = "Lai, Viet and Ngo, Nghia and Pouran Ben Veyseh, Amir and Man, Hieu and Dernoncourt, Franck and Bui, Trung and Nguyen, Thien", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.878", doi = "10.18653/v1/2023.findings-emnlp.878", pages = "13171--13189", abstract = "Over the last few years, large language models (LLMs) have emerged as the most important breakthroughs in natural language processing (NLP) that fundamentally transform research and developments in the field. ChatGPT represents one of the most exciting LLM systems developed recently to showcase impressive skills for language generation and highly attract public attention. Among various exciting applications discovered for ChatGPT in English, the model can process and generate texts for multiple languages due to its multilingual training data. Given the broad adoption of ChatGPT for English in different problems and areas, a natural question is whether ChatGPT can also be applied effectively for other languages or it is necessary to develop more language-specific technologies. The answer to this question requires a thorough evaluation of ChatGPT over multiple tasks with diverse languages and large datasets (i.e., beyond reported anecdotes), which is still missing or limited in current research. Our work aims to fill this gap for the evaluation of ChatGPT and similar LLMs to provide more comprehensive information for multilingual NLP applications. In particular, we evaluate ChatGPT on 7 different tasks, covering 37 diverse languages with high, medium, low, and extremely low resources. Compared to the performance of previous models, our extensive experiments demonstrate the worse performance of ChatGPT for different NLP tasks and languages, calling for further research to develop better models and understanding for multilingual learning.", }

Over the last few years, large language models (LLMs) have emerged as the most important breakthroughs in natural language processing (NLP) that fundamentally transform research and developments in the field. ChatGPT represents one of the most exciting LLM systems developed recently to showcase impressive skills for language generation and highly attract public attention. Among various exciting applications discovered for ChatGPT in English, the model can process and generate texts for multiple languages due to its multilingual training data. Given the broad adoption of ChatGPT for English in different problems and areas, a natural question is whether ChatGPT can also be applied effectively for other languages or it is necessary to develop more language-specific technologies. The answer to this question requires a thorough evaluation of ChatGPT over multiple tasks with diverse languages and large datasets (i.e., beyond reported anecdotes), which is still missing or limited in current research. Our work aims to fill this gap for the evaluation of ChatGPT and similar LLMs to provide more comprehensive information for multilingual NLP applications. In particular, we evaluate ChatGPT on 7 different tasks, covering 37 diverse languages with high, medium, low, and extremely low resources. Compared to the performance of previous models, our extensive experiments demonstrate the worse performance of ChatGPT for different NLP tasks and languages, calling for further research to develop better models and understanding for multilingual learning.

ChatGPT Beyond English: Towards a Comprehensive Evaluation of Large Language Models in Multilingual Learning

findings-emnlp.878

2304.05613

https://huggingface.co/papers/2304.05613

Poster

https://aclanthology.org/2023.findings-emnlp.879.bib

https://aclanthology.org/2023.findings-emnlp.879/

@inproceedings{muller-eberstein-etal-2023-subspace, title = "Subspace Chronicles: How Linguistic Information Emerges, Shifts and Interacts during Language Model Training", author = {M{\"u}ller-Eberstein, Max and van der Goot, Rob and Plank, Barbara and Titov, Ivan}, editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.879", doi = "10.18653/v1/2023.findings-emnlp.879", pages = "13190--13208", abstract = "Representational spaces learned via language modeling are fundamental to Natural Language Processing (NLP), however there has been limited understanding regarding how and when during training various types of linguistic information emerge and interact. Leveraging a novel information theoretic probing suite, which enables direct comparisons of not just task performance, but their representational subspaces, we analyze nine tasks covering syntax, semantics and reasoning, across 2M pre-training steps and five seeds. We identify critical learning phases across tasks and time, during which subspaces emerge, share information, and later disentangle to specialize. Across these phases, syntactic knowledge is acquired rapidly after 0.5{\%} of full training. Continued performance improvements primarily stem from the acquisition of open-domain knowledge, while semantics and reasoning tasks benefit from later boosts to long-range contextualization and higher specialization. Measuring cross-task similarity further reveals that linguistically related tasks share information throughout training, and do so more during the critical phase of learning than before or after. Our findings have implications for model interpretability, multi-task learning, and learning from limited data.", }

Representational spaces learned via language modeling are fundamental to Natural Language Processing (NLP), however there has been limited understanding regarding how and when during training various types of linguistic information emerge and interact. Leveraging a novel information theoretic probing suite, which enables direct comparisons of not just task performance, but their representational subspaces, we analyze nine tasks covering syntax, semantics and reasoning, across 2M pre-training steps and five seeds. We identify critical learning phases across tasks and time, during which subspaces emerge, share information, and later disentangle to specialize. Across these phases, syntactic knowledge is acquired rapidly after 0.5{\%} of full training. Continued performance improvements primarily stem from the acquisition of open-domain knowledge, while semantics and reasoning tasks benefit from later boosts to long-range contextualization and higher specialization. Measuring cross-task similarity further reveals that linguistically related tasks share information throughout training, and do so more during the critical phase of learning than before or after. Our findings have implications for model interpretability, multi-task learning, and learning from limited data.

Subspace Chronicles: How Linguistic Information Emerges, Shifts and Interacts during Language Model Training

findings-emnlp.879

2310.16484

https://huggingface.co/papers/2310.16484

Poster

https://aclanthology.org/2023.findings-emnlp.880.bib

https://aclanthology.org/2023.findings-emnlp.880/

@inproceedings{yang-etal-2023-demonstration, title = "Not All Demonstration Examples are Equally Beneficial: Reweighting Demonstration Examples for In-Context Learning", author = "Yang, Zhe and Dai, Damai and Wang, Peiyi and Sui, Zhifang", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.880", doi = "10.18653/v1/2023.findings-emnlp.880", pages = "13209--13221", abstract = "Large Language Models (LLMs) have recently gained the In-Context Learning (ICL) ability with the models scaling up, allowing them to quickly adapt to downstream tasks with only a few demonstration examples prepended in the input sequence. Nonetheless, the current practice of ICL treats all demonstration examples equally, which still warrants improvement, as the quality of examples is usually uneven. In this paper, we investigate how to determine approximately optimal weights for demonstration examples and how to apply them during ICL. To assess the quality of weights in the absence of additional validation data, we design a masked self-prediction (MSP) score that exhibits a strong correlation with the final ICL performance. To expedite the weight-searching process, we discretize the continuous weight space and adopt beam search. With approximately optimal weights obtained, we further propose two strategies to apply them to demonstrations at different model positions. Experimental results on 8 text classification tasks show that our approach outperforms conventional ICL by a large margin. Our code are publicly available at https:github.com/Zhe-Young/WICL.", }

Large Language Models (LLMs) have recently gained the In-Context Learning (ICL) ability with the models scaling up, allowing them to quickly adapt to downstream tasks with only a few demonstration examples prepended in the input sequence. Nonetheless, the current practice of ICL treats all demonstration examples equally, which still warrants improvement, as the quality of examples is usually uneven. In this paper, we investigate how to determine approximately optimal weights for demonstration examples and how to apply them during ICL. To assess the quality of weights in the absence of additional validation data, we design a masked self-prediction (MSP) score that exhibits a strong correlation with the final ICL performance. To expedite the weight-searching process, we discretize the continuous weight space and adopt beam search. With approximately optimal weights obtained, we further propose two strategies to apply them to demonstrations at different model positions. Experimental results on 8 text classification tasks show that our approach outperforms conventional ICL by a large margin. Our code are publicly available at https:github.com/Zhe-Young/WICL.

Not All Demonstration Examples are Equally Beneficial: Reweighting Demonstration Examples for In-Context Learning

findings-emnlp.880

2310.08309

Poster

https://aclanthology.org/2023.findings-emnlp.881.bib

https://aclanthology.org/2023.findings-emnlp.881/

@inproceedings{wilf-etal-2023-difference, title = "Difference-Masking: Choosing What to Mask in Continued Pretraining", author = "Wilf, Alex and Akter, Syeda and Mathur, Leena and Liang, Paul and Mathew, Sheryl and Shou, Mengrou and Nyberg, Eric and Morency, Louis-Philippe", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.881", doi = "10.18653/v1/2023.findings-emnlp.881", pages = "13222--13234", abstract = "The self-supervised objective of masked prediction has led to promising performance gains on a variety of downstream tasks. However, while most approaches randomly mask tokens, there is strong intuition that deciding what to mask can substantially improve learning outcomes. We investigate this in continued pretraining setting in which pretrained models continue to pretrain on domain-specific data before performing some downstream task. We introduce Difference-Masking, a masking strategy that automatically chooses what to mask during continued pretraining by considering what makes a task domain different from the pretraining domain. Empirically, we find that Difference-Masking outperforms baselines on continued pretraining settings across four diverse language-only and multimodal video tasks.", }

The self-supervised objective of masked prediction has led to promising performance gains on a variety of downstream tasks. However, while most approaches randomly mask tokens, there is strong intuition that deciding what to mask can substantially improve learning outcomes. We investigate this in continued pretraining setting in which pretrained models continue to pretrain on domain-specific data before performing some downstream task. We introduce Difference-Masking, a masking strategy that automatically chooses what to mask during continued pretraining by considering what makes a task domain different from the pretraining domain. Empirically, we find that Difference-Masking outperforms baselines on continued pretraining settings across four diverse language-only and multimodal video tasks.

Difference-Masking: Choosing What to Mask in Continued Pretraining

findings-emnlp.881

2305.14577

Poster

https://aclanthology.org/2023.findings-emnlp.882.bib

https://aclanthology.org/2023.findings-emnlp.882/

@inproceedings{saadi-etal-2023-learn, title = "Learn From One Specialized Sub-Teacher: One-to-One Mapping for Feature-Based Knowledge Distillation", author = "Saadi, Khouloud and Mitrovi{\'c}, Jelena and Granitzer, Michael", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.882", doi = "10.18653/v1/2023.findings-emnlp.882", pages = "13235--13245", abstract = "Knowledge distillation is known as an effective technique for compressing over-parameterized language models. In this work, we propose to break down the global feature distillation task into N local sub-tasks. In this new framework, we consider each neuron in the last hidden layer of the teacher network as a specialized sub-teacher. We also consider each neuron in the last hidden layer of the student network as a focused sub-student. We make each focused sub-student learn from one corresponding specialized sub-teacher and ignore the others. This will facilitate the task for the sub-student and keep it focused. Our proposed method is novel and can be combined with other distillation techniques. Empirical results show that our proposed approach outperforms the state-of-the-art methods by maintaining higher performance on most benchmark datasets. Furthermore, we propose a randomized variant of our approach, called Masked One-to-One Mapping. Rather than learning all the N sub-tasks simultaneously, we focus on learning a subset of these sub-tasks at each optimization step. This variant enables the student to digest the received flow of knowledge more effectively and yields superior results.", }

Knowledge distillation is known as an effective technique for compressing over-parameterized language models. In this work, we propose to break down the global feature distillation task into N local sub-tasks. In this new framework, we consider each neuron in the last hidden layer of the teacher network as a specialized sub-teacher. We also consider each neuron in the last hidden layer of the student network as a focused sub-student. We make each focused sub-student learn from one corresponding specialized sub-teacher and ignore the others. This will facilitate the task for the sub-student and keep it focused. Our proposed method is novel and can be combined with other distillation techniques. Empirical results show that our proposed approach outperforms the state-of-the-art methods by maintaining higher performance on most benchmark datasets. Furthermore, we propose a randomized variant of our approach, called Masked One-to-One Mapping. Rather than learning all the N sub-tasks simultaneously, we focus on learning a subset of these sub-tasks at each optimization step. This variant enables the student to digest the received flow of knowledge more effectively and yields superior results.

Learn From One Specialized Sub-Teacher: One-to-One Mapping for Feature-Based Knowledge Distillation

findings-emnlp.882

Poster

https://aclanthology.org/2023.findings-emnlp.883.bib

https://aclanthology.org/2023.findings-emnlp.883/

@inproceedings{moon-etal-2023-imu2clip, title = "{IMU}2{CLIP}: Language-grounded Motion Sensor Translation with Multimodal Contrastive Learning", author = "Moon, Seungwhan and Madotto, Andrea and Lin, Zhaojiang and Saraf, Aparajita and Bearman, Amy and Damavandi, Babak", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.883", doi = "10.18653/v1/2023.findings-emnlp.883", pages = "13246--13253", abstract = "We present IMU2CLIP, a novel pre-training approach to align Inertial Measurement Unit (IMU) motion sensor recordings with text and video, by projecting them into the joint representation space of Contrastive Language-Image Pre-training (CLIP). The proposed approach allows IMU2CLIP to translate human motions (as measured by IMU sensors) into their corresponding textual descriptions and videos {--} while preserving the transitivity across these modalities. We introduce several new IMU-based Wearable AI applications such as motion-based media search, or an LM-based multimodal reasoning with motion sensor data {--} all using text as the grounding platform. In addition, we show that IMU2CLIP significantly improves downstream performances when fine-tuned for each application, demonstrating its universal usage as a new pre-trained resource. Our code and models will be released publicly.", }

We present IMU2CLIP, a novel pre-training approach to align Inertial Measurement Unit (IMU) motion sensor recordings with text and video, by projecting them into the joint representation space of Contrastive Language-Image Pre-training (CLIP). The proposed approach allows IMU2CLIP to translate human motions (as measured by IMU sensors) into their corresponding textual descriptions and videos {--} while preserving the transitivity across these modalities. We introduce several new IMU-based Wearable AI applications such as motion-based media search, or an LM-based multimodal reasoning with motion sensor data {--} all using text as the grounding platform. In addition, we show that IMU2CLIP significantly improves downstream performances when fine-tuned for each application, demonstrating its universal usage as a new pre-trained resource. Our code and models will be released publicly.

IMU2CLIP: Language-grounded Motion Sensor Translation with Multimodal Contrastive Learning

findings-emnlp.883

Poster

https://aclanthology.org/2023.findings-emnlp.884.bib

https://aclanthology.org/2023.findings-emnlp.884/

@inproceedings{qu-etal-2023-conditioning, title = "Conditioning on Dialog Acts improves Empathy Style Transfer", author = "Qu, Renyi and Ungar, Lyle and Sedoc, Jo{\~a}o", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.884", doi = "10.18653/v1/2023.findings-emnlp.884", pages = "13254--13271", abstract = "We explore the role of dialog acts in style transfer, specifically empathy style transfer {--} rewriting a sentence to make it more empathetic without changing its meaning. Specifically, we use two novel few-shot prompting strategies: target prompting, which only uses examples of the target style (unlike traditional prompting with source/target pairs), and dialog-act-conditioned prompting, which first estimates the dialog act of the source sentence and then makes it more empathetic using few-shot examples of the same dialog act. Our study yields two key findings: (1) Target prompting typically improves empathy more effectively while maintaining the same level of semantic similarity; (2) Dialog acts matter. Dialog-act-conditioned prompting enhances empathy while preserving both semantics and the dialog-act type. Different dialog acts benefit differently from different prompting methods, highlighting the need for further investigation of the role of dialog acts in style transfer.", }

We explore the role of dialog acts in style transfer, specifically empathy style transfer {--} rewriting a sentence to make it more empathetic without changing its meaning. Specifically, we use two novel few-shot prompting strategies: target prompting, which only uses examples of the target style (unlike traditional prompting with source/target pairs), and dialog-act-conditioned prompting, which first estimates the dialog act of the source sentence and then makes it more empathetic using few-shot examples of the same dialog act. Our study yields two key findings: (1) Target prompting typically improves empathy more effectively while maintaining the same level of semantic similarity; (2) Dialog acts matter. Dialog-act-conditioned prompting enhances empathy while preserving both semantics and the dialog-act type. Different dialog acts benefit differently from different prompting methods, highlighting the need for further investigation of the role of dialog acts in style transfer.

Conditioning on Dialog Acts improves Empathy Style Transfer

findings-emnlp.884

Poster

https://aclanthology.org/2023.findings-emnlp.885.bib

https://aclanthology.org/2023.findings-emnlp.885/

@inproceedings{luo-etal-2023-systematic, title = "Systematic Assessment of Factual Knowledge in Large Language Models", author = "Luo, Linhao and Vu, Trang and Phung, Dinh and Haf, Reza", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.885", doi = "10.18653/v1/2023.findings-emnlp.885", pages = "13272--13286", abstract = "Previous studies have relied on existing question-answering benchmarks to evaluate the knowledge stored in large language models (LLMs). However, this approach has limitations regarding factual knowledge coverage, as it mostly focuses on generic domains which may overlap with the pretraining data. This paper proposes a framework to systematically assess the factual knowledge of LLMs by leveraging knowledge graphs (KGs). Our framework automatically generates a set of questions and expected answers from the facts stored in a given KG, and then evaluates the accuracy of LLMs in answering these questions. We systematically evaluate the state-of-the-art LLMs with KGs in generic and specific domains. The experiment shows that ChatGPT is consistently the top performer across all domains. We also find that LLMs performance depends on the instruction finetuning, domain and question complexity and is prone to adversarial context.", }

Previous studies have relied on existing question-answering benchmarks to evaluate the knowledge stored in large language models (LLMs). However, this approach has limitations regarding factual knowledge coverage, as it mostly focuses on generic domains which may overlap with the pretraining data. This paper proposes a framework to systematically assess the factual knowledge of LLMs by leveraging knowledge graphs (KGs). Our framework automatically generates a set of questions and expected answers from the facts stored in a given KG, and then evaluates the accuracy of LLMs in answering these questions. We systematically evaluate the state-of-the-art LLMs with KGs in generic and specific domains. The experiment shows that ChatGPT is consistently the top performer across all domains. We also find that LLMs performance depends on the instruction finetuning, domain and question complexity and is prone to adversarial context.

Systematic Assessment of Factual Knowledge in Large Language Models

findings-emnlp.885

2310.11638

Poster

https://aclanthology.org/2023.findings-emnlp.886.bib

https://aclanthology.org/2023.findings-emnlp.886/

@inproceedings{dong-etal-2023-speculation, title = "From Speculation Detection to Trustworthy Relational Tuples in Information Extraction", author = "Dong, Kuicai and Sun, Aixin and Kim, Jung-jae and Li, Xiaoli", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.886", doi = "10.18653/v1/2023.findings-emnlp.886", pages = "13287--13299", abstract = "Speculation detection is an important NLP task to identify text factuality. However, the extracted speculative information (e.g., speculative polarity, cue, and scope) lacks structure and poses challenges for direct utilization in downstream tasks. Open Information Extraction (OIE), on the other hand, extracts structured tuples as facts, without examining the certainty of these tuples. Bridging this gap between speculation detection and information extraction becomes imperative to generate structured speculative information and trustworthy relational tuples. Existing studies on speculation detection are defined at sentence level; but even if a sentence is determined to be speculative, not all factual tuples extracted from it are speculative. In this paper, we propose to study speculations in OIE tuples and determine whether a tuple is speculative. We formally define the research problem of tuple-level speculation detection. We then conduct detailed analysis on the LSOIE dataset which provides labels for speculative tuples. Lastly, we propose a baseline model SpecTup for this new research task.", }

Speculation detection is an important NLP task to identify text factuality. However, the extracted speculative information (e.g., speculative polarity, cue, and scope) lacks structure and poses challenges for direct utilization in downstream tasks. Open Information Extraction (OIE), on the other hand, extracts structured tuples as facts, without examining the certainty of these tuples. Bridging this gap between speculation detection and information extraction becomes imperative to generate structured speculative information and trustworthy relational tuples. Existing studies on speculation detection are defined at sentence level; but even if a sentence is determined to be speculative, not all factual tuples extracted from it are speculative. In this paper, we propose to study speculations in OIE tuples and determine whether a tuple is speculative. We formally define the research problem of tuple-level speculation detection. We then conduct detailed analysis on the LSOIE dataset which provides labels for speculative tuples. Lastly, we propose a baseline model SpecTup for this new research task.

From Speculation Detection to Trustworthy Relational Tuples in Information Extraction

findings-emnlp.886

Poster

https://aclanthology.org/2023.findings-emnlp.887.bib

https://aclanthology.org/2023.findings-emnlp.887/

@inproceedings{sun-etal-2023-tokenization, title = "Tokenization Consistency Matters for Generative Models on Extractive {NLP} Tasks", author = "Sun, Kaiser and Qi, Peng and Zhang, Yuhao and Liu, Lan and Wang, William and Huang, Zhiheng", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.887", doi = "10.18653/v1/2023.findings-emnlp.887", pages = "13300--13310", abstract = "Generative models have been widely applied to solve extractive tasks, where parts of the input is extracted to form the desired output, and achieved significant success. For example, in extractive question answering (QA), generative models have constantly yielded state-of-the-art results. In this work, we study the issue of tokenization inconsistency that is commonly neglected in training these models. This issue damages the extractive nature of these tasks after the input and output are tokenized inconsistently by the tokenizer, and thus leads to performance drop as well as hallucination. We propose a simple yet effective fix to this issue and conduct a case study on extractive QA. We show that, with consistent tokenization, the model performs better in both in-domain and out-of-domain datasets, with a notable average of +1.7 F1 gain when a BART model is trained on SQuAD and evaluated on 8 QA datasets. Further, the model converges faster, and becomes less likely to generate out-of-context answers. Our results demonstrate the need for increased scrutiny regarding how tokenization is done in extractive tasks and the benefits of consistent tokenization during training.", }

Generative models have been widely applied to solve extractive tasks, where parts of the input is extracted to form the desired output, and achieved significant success. For example, in extractive question answering (QA), generative models have constantly yielded state-of-the-art results. In this work, we study the issue of tokenization inconsistency that is commonly neglected in training these models. This issue damages the extractive nature of these tasks after the input and output are tokenized inconsistently by the tokenizer, and thus leads to performance drop as well as hallucination. We propose a simple yet effective fix to this issue and conduct a case study on extractive QA. We show that, with consistent tokenization, the model performs better in both in-domain and out-of-domain datasets, with a notable average of +1.7 F1 gain when a BART model is trained on SQuAD and evaluated on 8 QA datasets. Further, the model converges faster, and becomes less likely to generate out-of-context answers. Our results demonstrate the need for increased scrutiny regarding how tokenization is done in extractive tasks and the benefits of consistent tokenization during training.

Tokenization Consistency Matters for Generative Models on Extractive NLP Tasks

findings-emnlp.887

2212.09912

Poster

https://aclanthology.org/2023.findings-emnlp.888.bib

https://aclanthology.org/2023.findings-emnlp.888/

@inproceedings{gao-etal-2023-dialogue-medical, title = "Dialogue Medical Information Extraction with Medical-Item Graph and Dialogue-Status Enriched Representation", author = "Gao, Lei and Zhang, Xinnan and Wu, Xian and Ge, Shen and Zheng, Yefeng", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.888", doi = "10.18653/v1/2023.findings-emnlp.888", pages = "13311--13321", abstract = "The multi-turn doctor-patient dialogue includes rich medical knowledge, like the symptoms of the patient, the diagnosis and medication suggested by the doctor. If mined and represented properly, such medical knowledge can benefit a large range of clinical applications, including diagnosis assistance and medication recommendation. To derive structured knowledge from free text dialogues, we target a critical task: the Dialogue Medical Information Extraction (DMIE). DMIE aims to detect pre-defined clinical meaningful medical items (symptoms, surgery, etc.) as well as their statuses (positive, negative, etc.) from the dialogue. Existing approaches mainly formulate DMIE as a multi-label classification problem and ignore the relationships among medical items and statuses. Different from previous approaches, we propose a heterogeneous graph to model the relationship between items. We further propose two consecutive attention based modules to enrich the item representation with the dialogue and status. In this manner, we are able to model the relationships among medical items and statuses in the DMIE task. Experimental results on the public benchmark data set show that the proposed model outperforms previous works and achieves the state-of-the-art performance.", }

The multi-turn doctor-patient dialogue includes rich medical knowledge, like the symptoms of the patient, the diagnosis and medication suggested by the doctor. If mined and represented properly, such medical knowledge can benefit a large range of clinical applications, including diagnosis assistance and medication recommendation. To derive structured knowledge from free text dialogues, we target a critical task: the Dialogue Medical Information Extraction (DMIE). DMIE aims to detect pre-defined clinical meaningful medical items (symptoms, surgery, etc.) as well as their statuses (positive, negative, etc.) from the dialogue. Existing approaches mainly formulate DMIE as a multi-label classification problem and ignore the relationships among medical items and statuses. Different from previous approaches, we propose a heterogeneous graph to model the relationship between items. We further propose two consecutive attention based modules to enrich the item representation with the dialogue and status. In this manner, we are able to model the relationships among medical items and statuses in the DMIE task. Experimental results on the public benchmark data set show that the proposed model outperforms previous works and achieves the state-of-the-art performance.

Dialogue Medical Information Extraction with Medical-Item Graph and Dialogue-Status Enriched Representation

findings-emnlp.888

Poster

https://aclanthology.org/2023.findings-emnlp.889.bib

https://aclanthology.org/2023.findings-emnlp.889/

@inproceedings{nakamura-etal-2023-logicattack, title = "{L}ogic{A}ttack: Adversarial Attacks for Evaluating Logical Consistency of Natural Language Inference", author = "Nakamura, Mutsumi and Mashetty, Santosh and Parmar, Mihir and Varshney, Neeraj and Baral, Chitta", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.889", doi = "10.18653/v1/2023.findings-emnlp.889", pages = "13322--13334", abstract = "Recently Large Language Models (LLMs) such as GPT-3, ChatGPT, and FLAN have led to impressive progress in Natural Language Inference (NLI) tasks. However, these models may rely on simple heuristics or artifacts in the evaluation data to achieve their high performance, which suggests that they still suffer from logical inconsistency. To assess the logical consistency of these models, we propose a LogicAttack, a method to attack NLI models using diverse logical forms of premise and hypothesis, providing a more robust evaluation of their performance. Our approach leverages a range of inference rules from propositional logic, such as Modus Tollens and Bidirectional Dilemma, to generate effective adversarial attacks and identify common vulnerabilities across multiple NLI models. We achieve an average {\textasciitilde}53{\%} Attack Success Rate (ASR) across multiple logic-based attacks. Moreover, we demonstrate that incorporating generated attack samples into training enhances the logical reasoning ability of the target model and decreases its vulnerability to logic-based attacks. Data and source code are available at https://github.com/msantoshmadhav/LogicAttack.", }

Recently Large Language Models (LLMs) such as GPT-3, ChatGPT, and FLAN have led to impressive progress in Natural Language Inference (NLI) tasks. However, these models may rely on simple heuristics or artifacts in the evaluation data to achieve their high performance, which suggests that they still suffer from logical inconsistency. To assess the logical consistency of these models, we propose a LogicAttack, a method to attack NLI models using diverse logical forms of premise and hypothesis, providing a more robust evaluation of their performance. Our approach leverages a range of inference rules from propositional logic, such as Modus Tollens and Bidirectional Dilemma, to generate effective adversarial attacks and identify common vulnerabilities across multiple NLI models. We achieve an average {\textasciitilde}53{\%} Attack Success Rate (ASR) across multiple logic-based attacks. Moreover, we demonstrate that incorporating generated attack samples into training enhances the logical reasoning ability of the target model and decreases its vulnerability to logic-based attacks. Data and source code are available at https://github.com/msantoshmadhav/LogicAttack.

LogicAttack: Adversarial Attacks for Evaluating Logical Consistency of Natural Language Inference

findings-emnlp.889

Poster

https://aclanthology.org/2023.findings-emnlp.890.bib

https://aclanthology.org/2023.findings-emnlp.890/

@inproceedings{xiao-etal-2023-decomposed, title = "Decomposed Prompt Tuning via Low-Rank Reparameterization", author = "Xiao, Yao and Xu, Lu and Li, Jiaxi and Lu, Wei and Li, Xiaoli", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.890", doi = "10.18653/v1/2023.findings-emnlp.890", pages = "13335--13347", abstract = "While prompt tuning approaches have achieved competitive performance with high efficiency, we observe that they invariably employ the same initialization process, wherein the soft prompt is either randomly initialized or derived from an existing embedding vocabulary. In contrast to these conventional methods, this study aims to investigate an alternative way to derive soft prompt. Our empirical studies show that the soft prompt typically exhibits a low {``}intrinsic rank{''} characteristic. With such observations, we propose decomposed prompt tuning, a novel approach that utilizes low-rank matrices to initialize the soft prompt. Through the low-rank reparameterization, our method significantly reduces the number of trainable parameters while maintaining effectiveness. Experimental results on the SuperGLUE benchmark in both high-resource and low-resource scenarios demonstrate the effectiveness of the proposed method.", }

While prompt tuning approaches have achieved competitive performance with high efficiency, we observe that they invariably employ the same initialization process, wherein the soft prompt is either randomly initialized or derived from an existing embedding vocabulary. In contrast to these conventional methods, this study aims to investigate an alternative way to derive soft prompt. Our empirical studies show that the soft prompt typically exhibits a low {``}intrinsic rank{''} characteristic. With such observations, we propose decomposed prompt tuning, a novel approach that utilizes low-rank matrices to initialize the soft prompt. Through the low-rank reparameterization, our method significantly reduces the number of trainable parameters while maintaining effectiveness. Experimental results on the SuperGLUE benchmark in both high-resource and low-resource scenarios demonstrate the effectiveness of the proposed method.

Decomposed Prompt Tuning via Low-Rank Reparameterization

findings-emnlp.890

2310.10094

https://huggingface.co/papers/2310.10094

Poster

https://aclanthology.org/2023.findings-emnlp.891.bib

https://aclanthology.org/2023.findings-emnlp.891/

@inproceedings{zhang-etal-2023-sgp, title = "{SGP}-{TOD}: Building Task Bots Effortlessly via Schema-Guided {LLM} Prompting", author = "Zhang, Xiaoying and Peng, Baolin and Li, Kun and Zhou, Jingyan and Meng, Helen", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.891", doi = "10.18653/v1/2023.findings-emnlp.891", pages = "13348--13369", abstract = "Building and maintaining end-to-end task bots using minimal human effort is a long-standing challenge in dialog research. In this work, we introduce SGP-TOD, Schema-Guided Prompting for building Task-Oriented Dialog systems effortlessly based on large language models (LLMs). Utilizing the predefined task schema, i.e., belief instruction and dialog policy, we instruct fixed LLMs to generate appropriate responses on novel tasks, without the need for training data. Specifically, SGP-TOD comprises three components: an LLM for interacting with users, a Dialog State Tracking (DST) Prompter to aid the LLM in tracking dialog states with the given belief instruction, and a Policy Prompter to direct the LLM to generate proper responses adhering to the provided dialog policy. Experimental results on Multiwoz, RADDLE, and STAR datasets show that our training-free strategy, SGP-TOD, yields state-of-the-art (SOTA) zero-shot performance, significantly surpassing the few-shot approaches. In a domain-extension setting, SGP-TOD aptly adapts to new functionalities by merely adding supplementary schema rules. We make our code and data publicly available.", }

Building and maintaining end-to-end task bots using minimal human effort is a long-standing challenge in dialog research. In this work, we introduce SGP-TOD, Schema-Guided Prompting for building Task-Oriented Dialog systems effortlessly based on large language models (LLMs). Utilizing the predefined task schema, i.e., belief instruction and dialog policy, we instruct fixed LLMs to generate appropriate responses on novel tasks, without the need for training data. Specifically, SGP-TOD comprises three components: an LLM for interacting with users, a Dialog State Tracking (DST) Prompter to aid the LLM in tracking dialog states with the given belief instruction, and a Policy Prompter to direct the LLM to generate proper responses adhering to the provided dialog policy. Experimental results on Multiwoz, RADDLE, and STAR datasets show that our training-free strategy, SGP-TOD, yields state-of-the-art (SOTA) zero-shot performance, significantly surpassing the few-shot approaches. In a domain-extension setting, SGP-TOD aptly adapts to new functionalities by merely adding supplementary schema rules. We make our code and data publicly available.

SGP-TOD: Building Task Bots Effortlessly via Schema-Guided LLM Prompting

findings-emnlp.891

2305.09067

Poster

https://aclanthology.org/2023.findings-emnlp.892.bib

https://aclanthology.org/2023.findings-emnlp.892/

@inproceedings{rao-etal-2023-ethical, title = "Ethical Reasoning over Moral Alignment: A Case and Framework for In-Context Ethical Policies in {LLM}s", author = "Rao, Abhinav Sukumar and Khandelwal, Aditi and Tanmay, Kumar and Agarwal, Utkarsh and Choudhury, Monojit", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.892", doi = "10.18653/v1/2023.findings-emnlp.892", pages = "13370--13388", abstract = "In this position paper, we argue that instead of morally aligning LLMs to specific set of ethical principles, we should infuse generic ethical reasoning capabilities into them so that they can handle value pluralism at a global scale. When provided with an ethical policy, an LLM should be capable of making decisions that are ethically consistent to the policy. We develop a framework that integrates moral dilemmas with moral principles pertaining to different foramlisms of normative ethics, and at different levels of abstractions. Initial experiments with GPT-x models shows that while GPT-4 is a nearly perfect ethical reasoner, the models still have bias towards the moral values of Western and English speaking societies.", }

In this position paper, we argue that instead of morally aligning LLMs to specific set of ethical principles, we should infuse generic ethical reasoning capabilities into them so that they can handle value pluralism at a global scale. When provided with an ethical policy, an LLM should be capable of making decisions that are ethically consistent to the policy. We develop a framework that integrates moral dilemmas with moral principles pertaining to different foramlisms of normative ethics, and at different levels of abstractions. Initial experiments with GPT-x models shows that while GPT-4 is a nearly perfect ethical reasoner, the models still have bias towards the moral values of Western and English speaking societies.

Ethical Reasoning over Moral Alignment: A Case and Framework for In-Context Ethical Policies in LLMs

findings-emnlp.892

2310.07251

https://huggingface.co/papers/2310.07251

Poster

https://aclanthology.org/2023.findings-emnlp.893.bib

https://aclanthology.org/2023.findings-emnlp.893/

@inproceedings{luo-etal-2023-vector, title = "Vector-Quantized Prompt Learning for Paraphrase Generation", author = "Luo, Haotian and Liu, Yixin and Liu, Peidong and Liu, Xianggen", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.893", doi = "10.18653/v1/2023.findings-emnlp.893", pages = "13389--13398", abstract = "Deep generative modeling of natural languages has achieved many successes, such as producing fluent sentences and translating from one language into another. However, the development of generative modeling techniques for paraphrase generation still lags behind largely due to the challenges in addressing the complex conflicts between expression diversity and semantic preservation. This paper proposes to generate diverse and high-quality paraphrases by exploiting the pre-trained models with instance-dependent prompts. To learn generalizable prompts, we assume that the number of abstract transforming patterns of paraphrase generation (governed by prompts) is finite and usually not large. Therefore, we present vector-quantized prompts as the cues to control the generation of pre-trained models. Extensive experiments demonstrate that the proposed method achieves new state-of-art results on three benchmark datasets, including Quora, Wikianswers, and MSCOCO. We will release all the code upon acceptance.", }

Deep generative modeling of natural languages has achieved many successes, such as producing fluent sentences and translating from one language into another. However, the development of generative modeling techniques for paraphrase generation still lags behind largely due to the challenges in addressing the complex conflicts between expression diversity and semantic preservation. This paper proposes to generate diverse and high-quality paraphrases by exploiting the pre-trained models with instance-dependent prompts. To learn generalizable prompts, we assume that the number of abstract transforming patterns of paraphrase generation (governed by prompts) is finite and usually not large. Therefore, we present vector-quantized prompts as the cues to control the generation of pre-trained models. Extensive experiments demonstrate that the proposed method achieves new state-of-art results on three benchmark datasets, including Quora, Wikianswers, and MSCOCO. We will release all the code upon acceptance.

Vector-Quantized Prompt Learning for Paraphrase Generation

findings-emnlp.893

2311.14949

Poster

https://aclanthology.org/2023.findings-emnlp.894.bib

https://aclanthology.org/2023.findings-emnlp.894/

@inproceedings{li-etal-2023-rethinking-construction, title = "Rethinking the Construction of Effective Metrics for Understanding the Mechanisms of Pretrained Language Models", author = "Li, You and Yin, Jinhui and Lin, Yuming", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.894", doi = "10.18653/v1/2023.findings-emnlp.894", pages = "13399--13412", abstract = "Pretrained language models are expected to effectively map input text to a set of vectors while preserving the inherent relationships within the text. Consequently, designing a white-box model to compute metrics that reflect the presence of specific internal relations in these vectors has become a common approach for post-hoc interpretability analysis of pretrained language models. However, achieving interpretability in white-box models and ensuring the rigor of metric computation becomes challenging when the source model lacks inherent interpretability. Therefore, in this paper, we discuss striking a balance in this trade-off and propose a novel line to constructing metrics for understanding the mechanisms of pretrained language models. We have specifically designed a family of metrics along this line of investigation, and the model used to compute these metrics is referred to as the tree topological probe. We conducted measurements on BERT-large by using these metrics. Based on the experimental results, we propose a speculation regarding the working mechanism of BERT-like pretrained language models, as well as a strategy for enhancing fine-tuning performance by leveraging the topological probe to improve specific submodules.", }

Pretrained language models are expected to effectively map input text to a set of vectors while preserving the inherent relationships within the text. Consequently, designing a white-box model to compute metrics that reflect the presence of specific internal relations in these vectors has become a common approach for post-hoc interpretability analysis of pretrained language models. However, achieving interpretability in white-box models and ensuring the rigor of metric computation becomes challenging when the source model lacks inherent interpretability. Therefore, in this paper, we discuss striking a balance in this trade-off and propose a novel line to constructing metrics for understanding the mechanisms of pretrained language models. We have specifically designed a family of metrics along this line of investigation, and the model used to compute these metrics is referred to as the tree topological probe. We conducted measurements on BERT-large by using these metrics. Based on the experimental results, we propose a speculation regarding the working mechanism of BERT-like pretrained language models, as well as a strategy for enhancing fine-tuning performance by leveraging the topological probe to improve specific submodules.

Rethinking the Construction of Effective Metrics for Understanding the Mechanisms of Pretrained Language Models

findings-emnlp.894

2310.12454

Poster

https://aclanthology.org/2023.findings-emnlp.895.bib

https://aclanthology.org/2023.findings-emnlp.895/

@inproceedings{zhao-etal-2023-parrot, title = "{PARROT}: Zero-Shot Narrative Reading Comprehension via Parallel Reading", author = "Zhao, Chao and Vijjini, Anvesh and Chaturvedi, Snigdha", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.895", doi = "10.18653/v1/2023.findings-emnlp.895", pages = "13413--13424", abstract = "Narrative comprehension is a challenging task that requires a deep understanding of the foundational elements of narratives. Acquiring this skill requires extensive annotated data. To mitigate the burden of data annotation, we present Parrot, a zero-shot approach for narrative reading comprehension through parallel reading, which involves two parallel narratives that tell the same story. By leveraging one narrative as a source of supervision signal to guide the understanding of the other, Parrot abstracts the textual content and develops genuine narrative understanding. Evaluation conducted on two narrative comprehension benchmarks demonstrates that Parrot surpasses previous zero-shot approaches and achieves comparable performance to fully supervised models. The code will be available at https://github.com/zhaochaocs/Parrot.", }

Narrative comprehension is a challenging task that requires a deep understanding of the foundational elements of narratives. Acquiring this skill requires extensive annotated data. To mitigate the burden of data annotation, we present Parrot, a zero-shot approach for narrative reading comprehension through parallel reading, which involves two parallel narratives that tell the same story. By leveraging one narrative as a source of supervision signal to guide the understanding of the other, Parrot abstracts the textual content and develops genuine narrative understanding. Evaluation conducted on two narrative comprehension benchmarks demonstrates that Parrot surpasses previous zero-shot approaches and achieves comparable performance to fully supervised models. The code will be available at https://github.com/zhaochaocs/Parrot.

PARROT: Zero-Shot Narrative Reading Comprehension via Parallel Reading

findings-emnlp.895

Poster

https://aclanthology.org/2023.findings-emnlp.896.bib

https://aclanthology.org/2023.findings-emnlp.896/

@inproceedings{doosterlinck-etal-2023-biodex, title = "{B}io{DEX}: Large-Scale Biomedical Adverse Drug Event Extraction for Real-World Pharmacovigilance", author = "D{'}Oosterlinck, Karel and Remy, Fran{\c{c}}ois and Deleu, Johannes and Demeester, Thomas and Develder, Chris and Zaporojets, Klim and Ghodsi, Aneiss and Ellershaw, Simon and Collins, Jack and Potts, Christopher", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.896", doi = "10.18653/v1/2023.findings-emnlp.896", pages = "13425--13454", abstract = "Timely and accurate extraction of Adverse Drug Events (ADE) from biomedical literature is paramount for public safety, but involves slow and costly manual labor. We set out to improve drug safety monitoring (pharmacovigilance, PV) through the use of Natural Language Processing (NLP). We introduce BioDEX, a large-scale resource for Biomedical adverse Drug Event eXtraction, rooted in the historical output of drug safety reporting in the U.S. BioDEX consists of 65k abstracts and 19k full-text biomedical papers with 256k associated document-level safety reports created by medical experts. The core features of these reports include the reported weight, age, and biological sex of a patient, a set of drugs taken by the patient, the drug dosages, the reactions experienced, and whether the reaction was life threatening. In this work, we consider the task of predicting the core information of the report given its originating paper. We estimate human performance to be 72.0{\%} F1, whereas our best model achieves 59.1{\%} F1 (62.3 validation), indicating significant headroom. We also begin to explore ways in which these models could help professional PV reviewers. Our code and data are available at https://github.com/KarelDO/BioDEX.", }

Timely and accurate extraction of Adverse Drug Events (ADE) from biomedical literature is paramount for public safety, but involves slow and costly manual labor. We set out to improve drug safety monitoring (pharmacovigilance, PV) through the use of Natural Language Processing (NLP). We introduce BioDEX, a large-scale resource for Biomedical adverse Drug Event eXtraction, rooted in the historical output of drug safety reporting in the U.S. BioDEX consists of 65k abstracts and 19k full-text biomedical papers with 256k associated document-level safety reports created by medical experts. The core features of these reports include the reported weight, age, and biological sex of a patient, a set of drugs taken by the patient, the drug dosages, the reactions experienced, and whether the reaction was life threatening. In this work, we consider the task of predicting the core information of the report given its originating paper. We estimate human performance to be 72.0{\%} F1, whereas our best model achieves 59.1{\%} F1 (62.3 validation), indicating significant headroom. We also begin to explore ways in which these models could help professional PV reviewers. Our code and data are available at https://github.com/KarelDO/BioDEX.

BioDEX: Large-Scale Biomedical Adverse Drug Event Extraction for Real-World Pharmacovigilance

findings-emnlp.896

2305.13395

Poster

https://aclanthology.org/2023.findings-emnlp.897.bib

https://aclanthology.org/2023.findings-emnlp.897/

@inproceedings{an-etal-2023-coarse, title = "Coarse-to-Fine Dual Encoders are Better Frame Identification Learners", author = "An, Kaikai and Zheng, Ce and Gao, Bofei and Zhao, Haozhe and Chang, Baobao", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.897", doi = "10.18653/v1/2023.findings-emnlp.897", pages = "13455--13466", abstract = "Frame identification aims to find semantic frames associated with target words in a sentence. Recent researches measure the similarity or matching score between targets and candidate frames by modeling frame definitions. However, they either lack sufficient representation learning of the definitions or face challenges in efficiently selecting the most suitable frame from over 1000 candidate frames. Moreover, commonly used lexicon filtering ($lf$) to obtain candidate frames for the target may ignore out-of-vocabulary targets and cause inadequate frame modeling. In this paper, we propose CoFFTEA, a $\underline{Co}$arse-to-$\underline{F}$ine $\underline{F}$rame and $\underline{T}$arget $\underline{E}$ncoders $\underline{A}$rchitecture. With contrastive learning and dual encoders, CoFFTEA efficiently and effectively models the alignment between frames and targets. By employing a coarse-to-fine curriculum learning procedure, CoFFTEA gradually learns to differentiate frames with varying degrees of similarity. Experimental results demonstrate that CoFFTEA outperforms previous models by 0.93 overall scores and 1.53 R@1 without $lf$. Further analysis suggests that CoFFTEA can better model the relationships between frame and frame, as well as target and target. The code for our approach is available at https://github.com/pkunlp-icler/COFFTEA.", }

Frame identification aims to find semantic frames associated with target words in a sentence. Recent researches measure the similarity or matching score between targets and candidate frames by modeling frame definitions. However, they either lack sufficient representation learning of the definitions or face challenges in efficiently selecting the most suitable frame from over 1000 candidate frames. Moreover, commonly used lexicon filtering ($lf$) to obtain candidate frames for the target may ignore out-of-vocabulary targets and cause inadequate frame modeling. In this paper, we propose CoFFTEA, a $\underline{Co}$arse-to-$\underline{F}$ine $\underline{F}$rame and $\underline{T}$arget $\underline{E}$ncoders $\underline{A}$rchitecture. With contrastive learning and dual encoders, CoFFTEA efficiently and effectively models the alignment between frames and targets. By employing a coarse-to-fine curriculum learning procedure, CoFFTEA gradually learns to differentiate frames with varying degrees of similarity. Experimental results demonstrate that CoFFTEA outperforms previous models by 0.93 overall scores and 1.53 R@1 without $lf$. Further analysis suggests that CoFFTEA can better model the relationships between frame and frame, as well as target and target. The code for our approach is available at https://github.com/pkunlp-icler/COFFTEA.

Coarse-to-Fine Dual Encoders are Better Frame Identification Learners

findings-emnlp.897

2310.13316

Poster

https://aclanthology.org/2023.findings-emnlp.898.bib

https://aclanthology.org/2023.findings-emnlp.898/

@inproceedings{bae-etal-2023-sound, title = "Sound of Story: Multi-modal Storytelling with Audio", author = "Bae, Jaeyeon and Jeong, Seokhoon and Kang, Seokun and Han, Namgi and Lee, Jae-Yon and Kim, Hyounghun and Kim, Taehwan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.898", doi = "10.18653/v1/2023.findings-emnlp.898", pages = "13467--13479", abstract = "Storytelling is multi-modal in the real world. When one tells a story, one may use all of the visualizations and sounds along with the story itself. However, prior studies on storytelling datasets and tasks have paid little attention to sound even though sound also conveys meaningful semantics of the story. Therefore, we propose to extend story understanding and telling areas by establishing a new component called background sound which is story context-based audio without any linguistic information. For this purpose, we introduce a new dataset, called Sound of Story (SoS), which has paired image and text sequences with corresponding sound or background music for a story. To the best of our knowledge, this is the largest well-curated dataset for storytelling with sound. Our SoS dataset consists of 27,354 stories with 19.6 images per story and 984 hours of speech-decoupled audio such as background music and other sounds. As benchmark tasks for storytelling with sound and the dataset, we propose retrieval tasks between modalities, and audio generation tasks from image-text sequences, introducing strong baselines for them. We believe the proposed dataset and tasks may shed light on the multi-modal understanding of storytelling in terms of sound.", }

Storytelling is multi-modal in the real world. When one tells a story, one may use all of the visualizations and sounds along with the story itself. However, prior studies on storytelling datasets and tasks have paid little attention to sound even though sound also conveys meaningful semantics of the story. Therefore, we propose to extend story understanding and telling areas by establishing a new component called background sound which is story context-based audio without any linguistic information. For this purpose, we introduce a new dataset, called Sound of Story (SoS), which has paired image and text sequences with corresponding sound or background music for a story. To the best of our knowledge, this is the largest well-curated dataset for storytelling with sound. Our SoS dataset consists of 27,354 stories with 19.6 images per story and 984 hours of speech-decoupled audio such as background music and other sounds. As benchmark tasks for storytelling with sound and the dataset, we propose retrieval tasks between modalities, and audio generation tasks from image-text sequences, introducing strong baselines for them. We believe the proposed dataset and tasks may shed light on the multi-modal understanding of storytelling in terms of sound.

Sound of Story: Multi-modal Storytelling with Audio

findings-emnlp.898

2310.19264

Poster

https://aclanthology.org/2023.findings-emnlp.899.bib

https://aclanthology.org/2023.findings-emnlp.899/

@inproceedings{siledar-etal-2023-synthesize, title = "Synthesize, if you do not have: Effective Synthetic Dataset Creation Strategies for Self-Supervised Opinion Summarization in {E}-commerce", author = "Siledar, Tejpalsingh and Banerjee, Suman and Patil, Amey and Singh, Sudhanshu and Chelliah, Muthusamy and Garera, Nikesh and Bhattacharyya, Pushpak", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.899", doi = "10.18653/v1/2023.findings-emnlp.899", pages = "13480--13491", abstract = "In e-commerce, opinion summarization is the process of condensing the opinions presented in product reviews. However, the absence of large amounts of supervised datasets presents challenges in generating both aspect-specific and general opinion summaries. Existing approaches have attempted to address these challenges through synthetic dataset creation (SDC). However, general opinion summarization models struggle to generate summaries faithful to the input reviews whereas aspect-specific opinion summarization models are limited due to their reliance on human-specified aspects and seed words. To address this, we propose SDC strategies tailored for general and aspect-specific opinion summarization. We experimented on three e-commerce test sets: Oposum+, Amazon, and Flipkart. For general opinion summarization, pre-trained language model (PLM) fine-tuned on our general synthetic dataset surpass the SOTA on average by 2.3 R1 points. Faithfulness evaluation metrics and human evaluations indicate that our model-generated summaries are more faithful to the input compared to others. For aspect-specific opinion summarization, PLM fine-tuned on our aspect-specific synthetic dataset surpass SOTA by {\textasciitilde} 1 R1 point without the aid of any human-specified aspects or seed words.", }

In e-commerce, opinion summarization is the process of condensing the opinions presented in product reviews. However, the absence of large amounts of supervised datasets presents challenges in generating both aspect-specific and general opinion summaries. Existing approaches have attempted to address these challenges through synthetic dataset creation (SDC). However, general opinion summarization models struggle to generate summaries faithful to the input reviews whereas aspect-specific opinion summarization models are limited due to their reliance on human-specified aspects and seed words. To address this, we propose SDC strategies tailored for general and aspect-specific opinion summarization. We experimented on three e-commerce test sets: Oposum+, Amazon, and Flipkart. For general opinion summarization, pre-trained language model (PLM) fine-tuned on our general synthetic dataset surpass the SOTA on average by 2.3 R1 points. Faithfulness evaluation metrics and human evaluations indicate that our model-generated summaries are more faithful to the input compared to others. For aspect-specific opinion summarization, PLM fine-tuned on our aspect-specific synthetic dataset surpass SOTA by {\textasciitilde} 1 R1 point without the aid of any human-specified aspects or seed words.

Synthesize, if you do not have: Effective Synthetic Dataset Creation Strategies for Self-Supervised Opinion Summarization in E-commerce

findings-emnlp.899

Poster

https://aclanthology.org/2023.findings-emnlp.900.bib

https://aclanthology.org/2023.findings-emnlp.900/

@inproceedings{xu-etal-2023-leveraging, title = "Leveraging Contrastive Learning and Knowledge Distillation for Incomplete Modality Rumor Detection", author = "Xu, Fan and Fu, Pinyun and Huang, Qi and Zou, Bowei and Aw, AiTi and Wang, Mingwen", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.900", doi = "10.18653/v1/2023.findings-emnlp.900", pages = "13492--13503", abstract = "Rumors spread rapidly through online social microblogs at a relatively low cost, causing substantial economic losses and negative consequences in our daily lives. Existing rumor detection models often neglect the underlying semantic coherence between text and image components in multimodal posts, as well as the challenges posed by incomplete modalities in single modal posts, such as missing text or images. This paper presents CLKD-IMRD, a novel framework for Incomplete Modality Rumor Detection. CLKD-IMRD employs Contrastive Learning and Knowledge Distillation to capture the semantic consistency between text and image pairs, while also enhancing model generalization to incomplete modalities within individual posts. Extensive experimental results demonstrate that our CLKD-IMRD outperforms state-of-the-art methods on two English and two Chinese benchmark datasets for rumor detection in social media.", }

Rumors spread rapidly through online social microblogs at a relatively low cost, causing substantial economic losses and negative consequences in our daily lives. Existing rumor detection models often neglect the underlying semantic coherence between text and image components in multimodal posts, as well as the challenges posed by incomplete modalities in single modal posts, such as missing text or images. This paper presents CLKD-IMRD, a novel framework for Incomplete Modality Rumor Detection. CLKD-IMRD employs Contrastive Learning and Knowledge Distillation to capture the semantic consistency between text and image pairs, while also enhancing model generalization to incomplete modalities within individual posts. Extensive experimental results demonstrate that our CLKD-IMRD outperforms state-of-the-art methods on two English and two Chinese benchmark datasets for rumor detection in social media.

Leveraging Contrastive Learning and Knowledge Distillation for Incomplete Modality Rumor Detection

findings-emnlp.900

Poster

https://aclanthology.org/2023.findings-emnlp.901.bib

https://aclanthology.org/2023.findings-emnlp.901/

@inproceedings{yang-etal-2023-beyond, title = "Beyond Testers{'} Biases: Guiding Model Testing with Knowledge Bases using {LLM}s", author = "Yang, Chenyang and Rustogi, Rishabh and Brower-Sinning, Rachel and Lewis, Grace and Kaestner, Christian and Wu, Tongshuang", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.901", doi = "10.18653/v1/2023.findings-emnlp.901", pages = "13504--13519", abstract = "Current model testing work has mostly focused on creating test cases. Identifying what to test is a step that is largely ignored and poorly supported. We propose Weaver, an interactive tool that supports requirements elicitation for guiding model testing. Weaver uses large language models to generate knowledge bases and recommends concepts from them interactively, allowing testers to elicit requirements for further testing. Weaver provides rich external knowledge to testers and encourages testers to systematically explore diverse concepts beyond their own biases. In a user study, we show that both NLP experts and non-experts identified more, as well as more diverse concepts worth testing when using Weaver. Collectively, they found more than 200 failing test cases for stance detection with zero-shot ChatGPT. Our case studies further show that Weaver can help practitioners test models in real-world settings, where developers define more nuanced application scenarios (e.g., code understanding and transcript summarization) using LLMs.", }

Current model testing work has mostly focused on creating test cases. Identifying what to test is a step that is largely ignored and poorly supported. We propose Weaver, an interactive tool that supports requirements elicitation for guiding model testing. Weaver uses large language models to generate knowledge bases and recommends concepts from them interactively, allowing testers to elicit requirements for further testing. Weaver provides rich external knowledge to testers and encourages testers to systematically explore diverse concepts beyond their own biases. In a user study, we show that both NLP experts and non-experts identified more, as well as more diverse concepts worth testing when using Weaver. Collectively, they found more than 200 failing test cases for stance detection with zero-shot ChatGPT. Our case studies further show that Weaver can help practitioners test models in real-world settings, where developers define more nuanced application scenarios (e.g., code understanding and transcript summarization) using LLMs.

Beyond Testers' Biases: Guiding Model Testing with Knowledge Bases using LLMs

findings-emnlp.901

2310.09668

Poster

https://aclanthology.org/2023.findings-emnlp.902.bib

https://aclanthology.org/2023.findings-emnlp.902/

@inproceedings{wang-etal-2023-car, title = "{CAR}: Conceptualization-Augmented Reasoner for Zero-Shot Commonsense Question Answering", author = "Wang, Weiqi and Fang, Tianqing and Ding, Wenxuan and Xu, Baixuan and Liu, Xin and Song, Yangqiu and Bosselut, Antoine", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.902", doi = "10.18653/v1/2023.findings-emnlp.902", pages = "13520--13545", abstract = "The task of zero-shot commonsense question answering evaluates models on their capacity to reason about general scenarios beyond those presented in specific datasets. Existing approaches for tackling this task leverage external knowledge from CommonSense Knowledge Bases (CSKBs) by pre-training the model on synthetic QA pairs constructed from CSKBs. In these approaches, negative examples (distractors) are formulated by randomly sampling from CSKBs using fairly primitive keyword constraints. However, two bottlenecks limit these approaches: the inherent incompleteness of CSKBs limits the semantic coverage of synthetic QA pairs, and the lack of human annotations makes the sampled negative examples potentially uninformative and contradictory. To tackle these limitations above, we propose Conceptualization-Augmented Reasoner (CAR), a zero-shot commonsense question-answering framework that fully leverages the power of conceptualization. Specifically, CAR abstracts a commonsense knowledge triple to many higher-level instances, which increases the coverage of the CSKB and expands the ground-truth answer space, reducing the likelihood of selecting false negative distractors. Extensive experiments demonstrate that CAR more robustly generalizes to answering questions about zero-shot commonsense scenarios than existing methods, including large language models, such as GPT3.5 and ChatGPT. Our code, data, and model checkpoints are available at https://github.com/HKUST-KnowComp/CAR.", }

The task of zero-shot commonsense question answering evaluates models on their capacity to reason about general scenarios beyond those presented in specific datasets. Existing approaches for tackling this task leverage external knowledge from CommonSense Knowledge Bases (CSKBs) by pre-training the model on synthetic QA pairs constructed from CSKBs. In these approaches, negative examples (distractors) are formulated by randomly sampling from CSKBs using fairly primitive keyword constraints. However, two bottlenecks limit these approaches: the inherent incompleteness of CSKBs limits the semantic coverage of synthetic QA pairs, and the lack of human annotations makes the sampled negative examples potentially uninformative and contradictory. To tackle these limitations above, we propose Conceptualization-Augmented Reasoner (CAR), a zero-shot commonsense question-answering framework that fully leverages the power of conceptualization. Specifically, CAR abstracts a commonsense knowledge triple to many higher-level instances, which increases the coverage of the CSKB and expands the ground-truth answer space, reducing the likelihood of selecting false negative distractors. Extensive experiments demonstrate that CAR more robustly generalizes to answering questions about zero-shot commonsense scenarios than existing methods, including large language models, such as GPT3.5 and ChatGPT. Our code, data, and model checkpoints are available at https://github.com/HKUST-KnowComp/CAR.

CAR: Conceptualization-Augmented Reasoner for Zero-Shot Commonsense Question Answering

findings-emnlp.902

2305.14869

Poster

https://aclanthology.org/2023.findings-emnlp.903.bib

https://aclanthology.org/2023.findings-emnlp.903/

@inproceedings{bhardwaj-etal-2023-knn, title = "k{NN}-{CM}: A Non-parametric Inference-Phase Adaptation of Parametric Text Classifiers", author = "Bhardwaj, Rishabh and Li, Yingting and Majumder, Navonil and Cheng, Bo and Poria, Soujanya", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.903", doi = "10.18653/v1/2023.findings-emnlp.903", pages = "13546--13557", abstract = "Semi-parametric models exhibit the properties of both parametric and non-parametric modeling and have been shown to be effective in the next-word prediction language modeling task. However, there is a lack of studies on the text-discriminating properties of such models. We propose an inference-phase approach{---}\textit{k}-Nearest Neighbor Classification Model (\textit{k}NN-CM){---}that enhances the capacity of a pre-trained parametric text classifier by incorporating a simple neighborhood search through the representation space of (memorized) training samples. The final class prediction of \textit{k}NN-CM is based on the convex combination of probabilities obtained from \textit{k}NN search and prediction of the classifier. Our experiments show consistent performance improvements on eight SuperGLUE tasks, three adversarial natural language inference (ANLI) datasets, 11 question-answering (QA) datasets, and two sentiment classification datasets.", }

Semi-parametric models exhibit the properties of both parametric and non-parametric modeling and have been shown to be effective in the next-word prediction language modeling task. However, there is a lack of studies on the text-discriminating properties of such models. We propose an inference-phase approach{---}\textit{k}-Nearest Neighbor Classification Model (\textit{k}NN-CM){---}that enhances the capacity of a pre-trained parametric text classifier by incorporating a simple neighborhood search through the representation space of (memorized) training samples. The final class prediction of \textit{k}NN-CM is based on the convex combination of probabilities obtained from \textit{k}NN search and prediction of the classifier. Our experiments show consistent performance improvements on eight SuperGLUE tasks, three adversarial natural language inference (ANLI) datasets, 11 question-answering (QA) datasets, and two sentiment classification datasets.

kNN-CM: A Non-parametric Inference-Phase Adaptation of Parametric Text Classifiers

findings-emnlp.903

Poster

https://aclanthology.org/2023.findings-emnlp.904.bib

https://aclanthology.org/2023.findings-emnlp.904/

@inproceedings{ye-etal-2023-cross, title = "Cross-modality Data Augmentation for End-to-End Sign Language Translation", author = "Ye, Jinhui and Jiao, Wenxiang and Wang, Xing and Tu, Zhaopeng and Xiong, Hui", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.904", doi = "10.18653/v1/2023.findings-emnlp.904", pages = "13558--13571", abstract = "End-to-end sign language translation (SLT) aims to directly convert sign language videos into spoken language texts without intermediate representations. It has been challenging due to the data scarcity of labeled data and the modality gap between sign videos and texts. To tackle these challenges, we propose a novel Cross-modality Data Augmentation (XmDA) framework to transfer the powerful gloss-to-text translation capabilities to end-to-end sign language translation (i.e., video-to-text). Specifically, XmDA consists of two key components: cross-modality mix-up and cross-modality knowledge distillation. The former one explicitly encourages the alignment between sign video features and gloss embeddings to bridge the modality gap. The latter one utilizes the generation knowledge from gloss-to-text teacher models to guide the spoken language text generation. Experimental results on two widely used SLT datasets, i.e., PHOENIX-2014T and CSL-Daily, demonstrate that the proposed XmDA framework significantly and consistently outperforms the baseline models. Extensive analyses confirm our claim that XmDA enhances end-to-end sign language translation by reducing the representation distance between sign videos and glosses, as well as improving the translation of low-frequency words and long sentences.", }

End-to-end sign language translation (SLT) aims to directly convert sign language videos into spoken language texts without intermediate representations. It has been challenging due to the data scarcity of labeled data and the modality gap between sign videos and texts. To tackle these challenges, we propose a novel Cross-modality Data Augmentation (XmDA) framework to transfer the powerful gloss-to-text translation capabilities to end-to-end sign language translation (i.e., video-to-text). Specifically, XmDA consists of two key components: cross-modality mix-up and cross-modality knowledge distillation. The former one explicitly encourages the alignment between sign video features and gloss embeddings to bridge the modality gap. The latter one utilizes the generation knowledge from gloss-to-text teacher models to guide the spoken language text generation. Experimental results on two widely used SLT datasets, i.e., PHOENIX-2014T and CSL-Daily, demonstrate that the proposed XmDA framework significantly and consistently outperforms the baseline models. Extensive analyses confirm our claim that XmDA enhances end-to-end sign language translation by reducing the representation distance between sign videos and glosses, as well as improving the translation of low-frequency words and long sentences.

Cross-modality Data Augmentation for End-to-End Sign Language Translation

findings-emnlp.904

2305.11096

Poster

https://aclanthology.org/2023.findings-emnlp.905.bib

https://aclanthology.org/2023.findings-emnlp.905/

@inproceedings{lee-etal-2023-consistency, title = "Consistency is Key: On Data-Efficient Modality Transfer in Speech Translation", author = "Lee, Hojin and Lee, Changmin and Hwang, Seung-won", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.905", doi = "10.18653/v1/2023.findings-emnlp.905", pages = "13572--13581", abstract = "End-to-end approaches have shown promising results for speech translation (ST), but they suffer from its data scarcity compared to machine translation (MT). To address this, progressive training has become a common practice, of using external MT data during the fine-tuning phase. Despite of its prevalence and computational overhead, its validity is not extensively corroborated yet. This paper conducts an empirical investigation and finds that progressive training is ineffective. We identify learning-forgetting trade-off as a critical obstacle, then hypothesize and verify that consistency learning (CL) breaks the dilemma of learning-forgetting. The proposed method, which combines knowledge distillation (KD) and CL, outperforms the previous methods on MuST-C dataset even without additional data, and our proposed consistency-informed KD achieves additional improvements against KD+CL. Code and models are availble at https://github.com/hjlee1371/consistency-s2tt.", }

End-to-end approaches have shown promising results for speech translation (ST), but they suffer from its data scarcity compared to machine translation (MT). To address this, progressive training has become a common practice, of using external MT data during the fine-tuning phase. Despite of its prevalence and computational overhead, its validity is not extensively corroborated yet. This paper conducts an empirical investigation and finds that progressive training is ineffective. We identify learning-forgetting trade-off as a critical obstacle, then hypothesize and verify that consistency learning (CL) breaks the dilemma of learning-forgetting. The proposed method, which combines knowledge distillation (KD) and CL, outperforms the previous methods on MuST-C dataset even without additional data, and our proposed consistency-informed KD achieves additional improvements against KD+CL. Code and models are availble at https://github.com/hjlee1371/consistency-s2tt.

Consistency is Key: On Data-Efficient Modality Transfer in Speech Translation

findings-emnlp.905

Poster

https://aclanthology.org/2023.findings-emnlp.906.bib

https://aclanthology.org/2023.findings-emnlp.906/

@inproceedings{du-etal-2023-relation, title = "Relation-Aware Question Answering for Heterogeneous Knowledge Graphs", author = "Du, Haowei and Huang, Quzhe and Li, Chen and Zhang, Chen and Li, Yang and Zhao, Dongyan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.906", doi = "10.18653/v1/2023.findings-emnlp.906", pages = "13582--13592", abstract = "Multi-hop Knowledge Base Question Answering(KBQA) aims to find the answer entity in a knowledge graph (KG), which requires multiple steps of reasoning. Existing retrieval-based approaches solve this task by concentrating on the specific relation at different hops and predicting the intermediate entity within the reasoning path. However, these models fail to utilize information from head-tail entities and the semantic connection between relations to enhance the current relation representation, which undermines the information capturing of relations in KGs. To address this issue, we construct a \textbf{dual relation graph} where each node denotes a relation in the original KG (\textbf{primal entity graph}) and edges are constructed between relations sharing same head or tail entities. Then we iteratively do primal entity graph reasoning, dual relation graph information propagation, and interaction between these two graphs. In this way, the interaction between entity and relation is enhanced, and we derive better entity and relation representations. Experiments on two public datasets, WebQSP and CWQ, show that our approach achieves a significant performance gain over the prior state-of-the-art.", }

Multi-hop Knowledge Base Question Answering(KBQA) aims to find the answer entity in a knowledge graph (KG), which requires multiple steps of reasoning. Existing retrieval-based approaches solve this task by concentrating on the specific relation at different hops and predicting the intermediate entity within the reasoning path. However, these models fail to utilize information from head-tail entities and the semantic connection between relations to enhance the current relation representation, which undermines the information capturing of relations in KGs. To address this issue, we construct a \textbf{dual relation graph} where each node denotes a relation in the original KG (\textbf{primal entity graph}) and edges are constructed between relations sharing same head or tail entities. Then we iteratively do primal entity graph reasoning, dual relation graph information propagation, and interaction between these two graphs. In this way, the interaction between entity and relation is enhanced, and we derive better entity and relation representations. Experiments on two public datasets, WebQSP and CWQ, show that our approach achieves a significant performance gain over the prior state-of-the-art.

Relation-Aware Question Answering for Heterogeneous Knowledge Graphs

findings-emnlp.906

2312.11922

Poster

https://aclanthology.org/2023.findings-emnlp.907.bib

https://aclanthology.org/2023.findings-emnlp.907/

@inproceedings{yang-li-2023-instoptima, title = "{I}nst{O}ptima: Evolutionary Multi-objective Instruction Optimization via Large Language Model-based Instruction Operators", author = "Yang, Heng and Li, Ke", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.907", doi = "10.18653/v1/2023.findings-emnlp.907", pages = "13593--13602", abstract = "Instruction-based language modeling has received significant attention in pretrained language models. However, the efficiency of instruction engineering remains low and hinders the development of instruction studies. Recent studies have focused on automating instruction generation, but they primarily aim to improve performance without considering other crucial objectives that impact instruction quality, such as instruction length and perplexity. Therefore, we propose a novel approach (i.e., InstOptima) that treats instruction generation as an evolutionary multi-objective optimization problem. In contrast to text edition-based methods, our approach utilizes a large language model (LLM) to simulate instruction operators, including mutation and crossover. Furthermore, we introduce an objective-guided mechanism for these operators, allowing the LLM to comprehend the objectives and enhance the quality of the generated instructions. Experimental results demonstrate improved fine-tuning performance and the generation of a diverse set of high-quality instructions.", }

Instruction-based language modeling has received significant attention in pretrained language models. However, the efficiency of instruction engineering remains low and hinders the development of instruction studies. Recent studies have focused on automating instruction generation, but they primarily aim to improve performance without considering other crucial objectives that impact instruction quality, such as instruction length and perplexity. Therefore, we propose a novel approach (i.e., InstOptima) that treats instruction generation as an evolutionary multi-objective optimization problem. In contrast to text edition-based methods, our approach utilizes a large language model (LLM) to simulate instruction operators, including mutation and crossover. Furthermore, we introduce an objective-guided mechanism for these operators, allowing the LLM to comprehend the objectives and enhance the quality of the generated instructions. Experimental results demonstrate improved fine-tuning performance and the generation of a diverse set of high-quality instructions.

InstOptima: Evolutionary Multi-objective Instruction Optimization via Large Language Model-based Instruction Operators

findings-emnlp.907

2310.17630

Poster

https://aclanthology.org/2023.findings-emnlp.908.bib

https://aclanthology.org/2023.findings-emnlp.908/

@inproceedings{peng-etal-2023-less, title = "Less than One-shot: Named Entity Recognition via Extremely Weak Supervision", author = "Peng, Letian and Wang, Zihan and Shang, Jingbo", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.908", doi = "10.18653/v1/2023.findings-emnlp.908", pages = "13603--13616", abstract = "We study the named entity recognition (NER) problem under the extremely weak supervision (XWS) setting, where only one example entity per type is given in a context-free way. While one can see that XWS is \textit{lighter than one-shot} in terms of the amount of supervision, we propose a novel method X-NER that can outperform the state-of-the-art one-shot NER methods. We first mine entity spans that are similar to the example entities from an unlabelled training corpus. Instead of utilizing entity span representations from language models, we find it more effective to compare the context distributions before and after the span is replaced by the entity example. We then leverage the top-ranked spans as pseudo-labels to train an NER tagger. Extensive experiments and analyses on 4 NER datasets show the superior end-to-end NER performance of X-NER, outperforming the state-of-the-art few-shot methods with 1-shot supervision and ChatGPT annotations significantly. Finally, our X-NER possesses several notable properties, such as inheriting the cross-lingual abilities of the underlying language models.", }

We study the named entity recognition (NER) problem under the extremely weak supervision (XWS) setting, where only one example entity per type is given in a context-free way. While one can see that XWS is \textit{lighter than one-shot} in terms of the amount of supervision, we propose a novel method X-NER that can outperform the state-of-the-art one-shot NER methods. We first mine entity spans that are similar to the example entities from an unlabelled training corpus. Instead of utilizing entity span representations from language models, we find it more effective to compare the context distributions before and after the span is replaced by the entity example. We then leverage the top-ranked spans as pseudo-labels to train an NER tagger. Extensive experiments and analyses on 4 NER datasets show the superior end-to-end NER performance of X-NER, outperforming the state-of-the-art few-shot methods with 1-shot supervision and ChatGPT annotations significantly. Finally, our X-NER possesses several notable properties, such as inheriting the cross-lingual abilities of the underlying language models.

Less than One-shot: Named Entity Recognition via Extremely Weak Supervision

findings-emnlp.908

2311.02861

Poster

https://aclanthology.org/2023.findings-emnlp.909.bib

https://aclanthology.org/2023.findings-emnlp.909/

@inproceedings{yun-etal-2023-focus, title = "Focus on the Core: Efficient Attention via Pruned Token Compression for Document Classification", author = "Yun, Jungmin and Kim, Mihyeon and Kim, Youngbin", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.909", doi = "10.18653/v1/2023.findings-emnlp.909", pages = "13617--13628", abstract = "Transformer-based models have achieved dominant performance in numerous NLP tasks. Despite their remarkable successes, pre-trained transformers such as BERT suffer from a computationally expensive self-attention mechanism that interacts with all tokens, including the ones unfavorable to classification performance. To overcome these challenges, we propose integrating two strategies: token pruning and token combining. Token pruning eliminates less important tokens in the attention mechanism{'}s key and value as they pass through the layers. Additionally, we adopt fuzzy logic to handle uncertainty and alleviate potential mispruning risks arising from an imbalanced distribution of each token{'}s importance. Token combining, on the other hand, condenses input sequences into smaller sizes in order to further compress the model. By integrating these two approaches, we not only improve the model{'}s performance but also reduce its computational demands. Experiments with various datasets demonstrate superior performance compared to baseline models, especially with the best improvement over the existing BERT model, achieving +5{\%}p in accuracy and +5.6{\%}p in F1 score. Additionally, memory cost is reduced to 0.61x, and a speedup of 1.64x is achieved.", }

Transformer-based models have achieved dominant performance in numerous NLP tasks. Despite their remarkable successes, pre-trained transformers such as BERT suffer from a computationally expensive self-attention mechanism that interacts with all tokens, including the ones unfavorable to classification performance. To overcome these challenges, we propose integrating two strategies: token pruning and token combining. Token pruning eliminates less important tokens in the attention mechanism{'}s key and value as they pass through the layers. Additionally, we adopt fuzzy logic to handle uncertainty and alleviate potential mispruning risks arising from an imbalanced distribution of each token{'}s importance. Token combining, on the other hand, condenses input sequences into smaller sizes in order to further compress the model. By integrating these two approaches, we not only improve the model{'}s performance but also reduce its computational demands. Experiments with various datasets demonstrate superior performance compared to baseline models, especially with the best improvement over the existing BERT model, achieving +5{\%}p in accuracy and +5.6{\%}p in F1 score. Additionally, memory cost is reduced to 0.61x, and a speedup of 1.64x is achieved.

Focus on the Core: Efficient Attention via Pruned Token Compression for Document Classification

findings-emnlp.909

2406.01283

Poster

https://aclanthology.org/2023.findings-emnlp.910.bib

https://aclanthology.org/2023.findings-emnlp.910/

@inproceedings{eyal-etal-2023-semantic, title = "Semantic Decomposition of Question and {SQL} for Text-to-{SQL} Parsing", author = "Eyal, Ben and Mahabi, Moran and Haroche, Ophir and Bachar, Amir and Elhadad, Michael", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.910", doi = "10.18653/v1/2023.findings-emnlp.910", pages = "13629--13645", abstract = "Text-to-SQL semantic parsing faces challenges in generalizing to cross-domain and complex queries. Recent research has employed a question decomposition strategy to enhance the parsing of complex SQL queries.However, this strategy encounters two major obstacles: (1) existing datasets lack question decomposition; (2) due to the syntactic complexity of SQL, most complex queries cannot be disentangled into sub-queries that can be readily recomposed. To address these challenges, we propose a new modular Query Plan Language (QPL) that systematically decomposes SQL queries into simple and regular sub-queries. We develop a translator from SQL to QPL by leveraging analysis of SQL server query optimization plans, and we augment the Spider dataset with QPL programs. Experimental results demonstrate that the modular nature of QPL benefits existing semantic-parsing architectures, and training text-to-QPL parsers is more effective than text-to-SQL parsing for semantically equivalent queries. The QPL approach offers two additional advantages: (1) QPL programs can be paraphrased as simple questions, which allows us to create a dataset of (complex question, decomposed questions). Training on this dataset, we obtain a Question Decomposer for data retrieval that is sensitive to database schemas. (2) QPL is more accessible to non-experts for complex queries, leading to more interpretable output from the semantic parser.", }

Text-to-SQL semantic parsing faces challenges in generalizing to cross-domain and complex queries. Recent research has employed a question decomposition strategy to enhance the parsing of complex SQL queries.However, this strategy encounters two major obstacles: (1) existing datasets lack question decomposition; (2) due to the syntactic complexity of SQL, most complex queries cannot be disentangled into sub-queries that can be readily recomposed. To address these challenges, we propose a new modular Query Plan Language (QPL) that systematically decomposes SQL queries into simple and regular sub-queries. We develop a translator from SQL to QPL by leveraging analysis of SQL server query optimization plans, and we augment the Spider dataset with QPL programs. Experimental results demonstrate that the modular nature of QPL benefits existing semantic-parsing architectures, and training text-to-QPL parsers is more effective than text-to-SQL parsing for semantically equivalent queries. The QPL approach offers two additional advantages: (1) QPL programs can be paraphrased as simple questions, which allows us to create a dataset of (complex question, decomposed questions). Training on this dataset, we obtain a Question Decomposer for data retrieval that is sensitive to database schemas. (2) QPL is more accessible to non-experts for complex queries, leading to more interpretable output from the semantic parser.

Semantic Decomposition of Question and SQL for Text-to-SQL Parsing

findings-emnlp.910

2310.13575

https://huggingface.co/papers/2310.13575

Poster

https://aclanthology.org/2023.findings-emnlp.911.bib

https://aclanthology.org/2023.findings-emnlp.911/

@inproceedings{ren-etal-2023-time, title = "Time-Aware Language Modeling for Historical Text Dating", author = "Ren, Han and Wang, Hai and Zhao, Yajie and Ren, Yafeng", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.911", doi = "10.18653/v1/2023.findings-emnlp.911", pages = "13646--13656", abstract = "Automatic text dating(ATD) is a challenging task since explicit temporal mentions usually do not appear in texts. Existing state-of-the-art approaches learn word representations via language models, whereas most of them ignore diachronic change of words, which may affect the efforts of text modeling. Meanwhile, few of them consider text modeling for long diachronic documents. In this paper, we present a time-aware language model named TALM, to learn temporal word representations by transferring language models of general domains to those of time-specific ones. We also build a hierarchical modeling approach to represent diachronic documents by encoding them with temporal word representations. Experiments on a Chinese diachronic corpus show that our model effectively captures implicit temporal information of words, and outperforms state-of-the-art approaches in historical text dating as well.", }

Automatic text dating(ATD) is a challenging task since explicit temporal mentions usually do not appear in texts. Existing state-of-the-art approaches learn word representations via language models, whereas most of them ignore diachronic change of words, which may affect the efforts of text modeling. Meanwhile, few of them consider text modeling for long diachronic documents. In this paper, we present a time-aware language model named TALM, to learn temporal word representations by transferring language models of general domains to those of time-specific ones. We also build a hierarchical modeling approach to represent diachronic documents by encoding them with temporal word representations. Experiments on a Chinese diachronic corpus show that our model effectively captures implicit temporal information of words, and outperforms state-of-the-art approaches in historical text dating as well.

Time-Aware Language Modeling for Historical Text Dating

findings-emnlp.911

Poster

https://aclanthology.org/2023.findings-emnlp.912.bib

https://aclanthology.org/2023.findings-emnlp.912/

@inproceedings{xu-etal-2023-read, title = "A Read-and-Select Framework for Zero-shot Entity Linking", author = "Xu, Zhenran and Chen, Yulin and Hu, Baotian and Zhang, Min", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.912", doi = "10.18653/v1/2023.findings-emnlp.912", pages = "13657--13666", abstract = "Zero-shot entity linking (EL) aims at aligning entity mentions to unseen entities to challenge the generalization ability. Previous methods largely focus on the candidate retrieval stage and ignore the essential candidate ranking stage, which disambiguates among entities and makes the final linking prediction. In this paper, we propose a read-and-select (ReS) framework by modeling the main components of entity disambiguation, i.e., mention-entity matching and cross-entity comparison. First, for each candidate, the reading module leverages mention context to output mention-aware entity representations, enabling mention-entity matching. Then, in the selecting module, we frame the choice of candidates as a sequence labeling problem, and all candidate representations are fused together to enable cross-entity comparison. Our method achieves the state-of-the-art performance on the established zero-shot EL dataset ZESHEL with a 2.55{\%} micro-average accuracy gain, with no need for laborious multi-phase pre-training used in most of the previous work, showing the effectiveness of both mention-entity and cross-entity interaction.", }

Zero-shot entity linking (EL) aims at aligning entity mentions to unseen entities to challenge the generalization ability. Previous methods largely focus on the candidate retrieval stage and ignore the essential candidate ranking stage, which disambiguates among entities and makes the final linking prediction. In this paper, we propose a read-and-select (ReS) framework by modeling the main components of entity disambiguation, i.e., mention-entity matching and cross-entity comparison. First, for each candidate, the reading module leverages mention context to output mention-aware entity representations, enabling mention-entity matching. Then, in the selecting module, we frame the choice of candidates as a sequence labeling problem, and all candidate representations are fused together to enable cross-entity comparison. Our method achieves the state-of-the-art performance on the established zero-shot EL dataset ZESHEL with a 2.55{\%} micro-average accuracy gain, with no need for laborious multi-phase pre-training used in most of the previous work, showing the effectiveness of both mention-entity and cross-entity interaction.

A Read-and-Select Framework for Zero-shot Entity Linking

findings-emnlp.912

2310.12450

Poster

https://aclanthology.org/2023.findings-emnlp.913.bib

https://aclanthology.org/2023.findings-emnlp.913/

@inproceedings{kongyoung-etal-2023-multi, title = "Multi-Task Learning of Query Generation and Classification for Generative Conversational Question Rewriting", author = "Kongyoung, Sarawoot and MacDonald, Craig and Ounis, Iadh", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.913", doi = "10.18653/v1/2023.findings-emnlp.913", pages = "13667--13678", abstract = "In conversational search settings, users ask questions and receive answers as part of a conversation. The ambiguity in the questions is a common challenge, which can be effectively addressed by leveraging contextual information from the conversation history. In this context, determining topic continuity and reformulating questions into well-defined queries are crucial tasks. Previous approaches have typically addressed these tasks either as a classification task in the case of topic continuity or as a text generation task for question reformulation. However, no prior work has combined both tasks to effectively identify ambiguous questions as part of a conversation. In this paper, we propose a Multi-Task Learning (MTL) approach that uses a text generation model for both question rewriting and classification. Our models, based on BART and T5, are trained to rewrite conversational questions and identify follow-up questions simultaneously. We evaluate our approach on multiple test sets and demonstrate that it outperforms single-task learning baselines on the three LIF test sets, with statistically significant improvements ranging from +3.5{\%} to +10.5{\%} in terms of F1 and Micro-F1 scores. We also show that our approach outperforms single-task question rewriting models in passage retrieval on a large OR-QuAC test set.", }

In conversational search settings, users ask questions and receive answers as part of a conversation. The ambiguity in the questions is a common challenge, which can be effectively addressed by leveraging contextual information from the conversation history. In this context, determining topic continuity and reformulating questions into well-defined queries are crucial tasks. Previous approaches have typically addressed these tasks either as a classification task in the case of topic continuity or as a text generation task for question reformulation. However, no prior work has combined both tasks to effectively identify ambiguous questions as part of a conversation. In this paper, we propose a Multi-Task Learning (MTL) approach that uses a text generation model for both question rewriting and classification. Our models, based on BART and T5, are trained to rewrite conversational questions and identify follow-up questions simultaneously. We evaluate our approach on multiple test sets and demonstrate that it outperforms single-task learning baselines on the three LIF test sets, with statistically significant improvements ranging from +3.5{\%} to +10.5{\%} in terms of F1 and Micro-F1 scores. We also show that our approach outperforms single-task question rewriting models in passage retrieval on a large OR-QuAC test set.

Multi-Task Learning of Query Generation and Classification for Generative Conversational Question Rewriting

findings-emnlp.913

Poster

https://aclanthology.org/2023.findings-emnlp.914.bib

https://aclanthology.org/2023.findings-emnlp.914/

@inproceedings{sandhan-etal-2023-depnecti, title = "{D}ep{N}e{CTI}: Dependency-based Nested Compound Type Identification for {S}anskrit", author = "Sandhan, Jivnesh and Narsupalli, Yaswanth and Muppirala, Sreevatsa and Krishnan, Sriram and Satuluri, Pavankumar and Kulkarni, Amba and Goyal, Pawan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.914", doi = "10.18653/v1/2023.findings-emnlp.914", pages = "13679--13692", abstract = "Multi-component compounding is a prevalent phenomenon in Sanskrit, and understanding the implicit structure of a compound{'}s components is crucial for deciphering its meaning. Earlier approaches in Sanskrit have focused on binary compounds and neglected the multi-component compound setting. This work introduces the novel task of nested compound type identification (NeCTI), which aims to identify nested spans of a multi-component compound and decode the implicit semantic relations between them. To the best of our knowledge, this is the first attempt in the field of lexical semantics to propose this task. We present 2 newly annotated datasets including an out-of-domain dataset for this task. We also benchmark these datasets by exploring the efficacy of the standard problem formulations such as nested named entity recognition, constituency parsing and seq2seq, etc. We present a novel framework named DepNeCTI: Dependency-based Nested Compound Type Identifier that surpasses the performance of the best baseline with an average absolute improvement of 13.1 points F1-score in terms of Labeled Span Score (LSS) and a 5-fold enhancement in inference efficiency. In line with the previous findings in the binary Sanskrit compound identification task, context provides benefits for the NeCTI task. The codebase and datasets are publicly available at: https://github.com/yaswanth-iitkgp/DepNeCTI", }

Multi-component compounding is a prevalent phenomenon in Sanskrit, and understanding the implicit structure of a compound{'}s components is crucial for deciphering its meaning. Earlier approaches in Sanskrit have focused on binary compounds and neglected the multi-component compound setting. This work introduces the novel task of nested compound type identification (NeCTI), which aims to identify nested spans of a multi-component compound and decode the implicit semantic relations between them. To the best of our knowledge, this is the first attempt in the field of lexical semantics to propose this task. We present 2 newly annotated datasets including an out-of-domain dataset for this task. We also benchmark these datasets by exploring the efficacy of the standard problem formulations such as nested named entity recognition, constituency parsing and seq2seq, etc. We present a novel framework named DepNeCTI: Dependency-based Nested Compound Type Identifier that surpasses the performance of the best baseline with an average absolute improvement of 13.1 points F1-score in terms of Labeled Span Score (LSS) and a 5-fold enhancement in inference efficiency. In line with the previous findings in the binary Sanskrit compound identification task, context provides benefits for the NeCTI task. The codebase and datasets are publicly available at: https://github.com/yaswanth-iitkgp/DepNeCTI

DepNeCTI: Dependency-based Nested Compound Type Identification for Sanskrit

findings-emnlp.914

2310.09501

Poster

https://aclanthology.org/2023.findings-emnlp.915.bib

https://aclanthology.org/2023.findings-emnlp.915/

@inproceedings{cohen-etal-2023-heq, title = "{H}e{Q}: a Large and Diverse {H}ebrew Reading Comprehension Benchmark", author = "Cohen, Amir and Merhav-Fine, Hilla and Goldberg, Yoav and Tsarfaty, Reut", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.915", doi = "10.18653/v1/2023.findings-emnlp.915", pages = "13693--13705", abstract = "Current benchmarks for Hebrew Natural Language Processing (NLP) focus mainly on morpho-syntactic tasks, neglecting the semantic dimension of language understanding. To bridge this gap, we set out to deliver a Hebrew Machine Reading Comprehension (MRC) dataset, where MRC is to be realized as extractive Question Answering. The morphologically-rich nature of Hebrew poses a challenge to this endeavor: the indeterminacy and non-transparency of span boundaries in morphologically complex forms lead to annotation inconsistencies, disagreements, and flaws of standard evaluation metrics. To remedy this, we devise a novel set of guidelines, a controlled crowdsourcing protocol, and revised evaluation metrics, that are suitable for the morphologically rich nature of the language. Our resulting benchmark, HeQ (Hebrew QA), features 30,147 diverse question-answer pairs derived from both Hebrew Wikipedia articles and Israeli tech news. Our empirical investigation reveals that standard evaluation metrics such as F1 Scores and Exact Match (EM) are not appropriate for Hebrew (and other MRLs), and we propose a relevant enhancement. In addition, our experiments show low correlation between models{'} performance on morpho-syntactic tasks and on MRC, which suggests that models that are designed for the former might underperform on semantic-heavy tasks. The development and exploration of HeQ illustrate some of the challenges MRLs pose in natural language understanding (NLU), fostering progression towards more and better NLU models for Hebrew and other MRLs.", }

Current benchmarks for Hebrew Natural Language Processing (NLP) focus mainly on morpho-syntactic tasks, neglecting the semantic dimension of language understanding. To bridge this gap, we set out to deliver a Hebrew Machine Reading Comprehension (MRC) dataset, where MRC is to be realized as extractive Question Answering. The morphologically-rich nature of Hebrew poses a challenge to this endeavor: the indeterminacy and non-transparency of span boundaries in morphologically complex forms lead to annotation inconsistencies, disagreements, and flaws of standard evaluation metrics. To remedy this, we devise a novel set of guidelines, a controlled crowdsourcing protocol, and revised evaluation metrics, that are suitable for the morphologically rich nature of the language. Our resulting benchmark, HeQ (Hebrew QA), features 30,147 diverse question-answer pairs derived from both Hebrew Wikipedia articles and Israeli tech news. Our empirical investigation reveals that standard evaluation metrics such as F1 Scores and Exact Match (EM) are not appropriate for Hebrew (and other MRLs), and we propose a relevant enhancement. In addition, our experiments show low correlation between models{'} performance on morpho-syntactic tasks and on MRC, which suggests that models that are designed for the former might underperform on semantic-heavy tasks. The development and exploration of HeQ illustrate some of the challenges MRLs pose in natural language understanding (NLU), fostering progression towards more and better NLU models for Hebrew and other MRLs.

HeQ: a Large and Diverse Hebrew Reading Comprehension Benchmark

findings-emnlp.915

Poster

https://aclanthology.org/2023.findings-emnlp.916.bib

https://aclanthology.org/2023.findings-emnlp.916/

@inproceedings{tripto-etal-2023-hansen, title = "{HANSEN}: Human and {AI} Spoken Text Benchmark for Authorship Analysis", author = "Tripto, Nafis and Uchendu, Adaku and Le, Thai and Setzu, Mattia and Giannotti, Fosca and Lee, Dongwon", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.916", doi = "10.18653/v1/2023.findings-emnlp.916", pages = "13706--13724", abstract = "$\textit{Authorship Analysis}$, also known as stylometry, has been an essential aspect of Natural Language Processing (NLP) for a long time. Likewise, the recent advancement of Large Language Models (LLMs) has made authorship analysis increasingly crucial for distinguishing between human-written and AI-generated texts. However, these authorship analysis tasks have primarily been focused on $\textit{written texts}$, not considering $\textit{spoken texts}$. Thus, we introduce the largest benchmark for spoken texts - ${\sf HANSEN}$($\underline{H}$uman $\underline{AN}$d ai $\underline{S}$poken t$\underline{E}$xt be$\underline{N}$chmark). ${\sf HANSEN}$ encompasses meticulous curation of existing speech datasets accompanied by transcripts, alongside the creation of novel AI-generated spoken text datasets. Together, it comprises 17 human datasets, and AI-generated spoken texts created using 3 prominent LLMs: ChatGPT, PaLM2, and Vicuna13B. To evaluate and demonstrate the utility of ${\sf HANSEN}$, we perform Authorship Attribution (AA) {\&} Author Verification (AV) on human-spoken datasets and conducted Human vs. AI text detection using state-of-the-art (SOTA) models. While SOTA methods, such as, character n-gram or Transformer-based model, exhibit similar AA {\&} AV performance in human-spoken datasets compared to written ones, there is much room for improvement in AI-generated spoken text detection. The ${\sf HANSEN}$ benchmark is available at: https://huggingface.co/datasets/HANSEN-REPO/HANSEN", }

$\textit{Authorship Analysis}$, also known as stylometry, has been an essential aspect of Natural Language Processing (NLP) for a long time. Likewise, the recent advancement of Large Language Models (LLMs) has made authorship analysis increasingly crucial for distinguishing between human-written and AI-generated texts. However, these authorship analysis tasks have primarily been focused on $\textit{written texts}$, not considering $\textit{spoken texts}$. Thus, we introduce the largest benchmark for spoken texts - ${\sf HANSEN}$($\underline{H}$uman $\underline{AN}$d ai $\underline{S}$poken t$\underline{E}$xt be$\underline{N}$chmark). ${\sf HANSEN}$ encompasses meticulous curation of existing speech datasets accompanied by transcripts, alongside the creation of novel AI-generated spoken text datasets. Together, it comprises 17 human datasets, and AI-generated spoken texts created using 3 prominent LLMs: ChatGPT, PaLM2, and Vicuna13B. To evaluate and demonstrate the utility of ${\sf HANSEN}$, we perform Authorship Attribution (AA) {\&} Author Verification (AV) on human-spoken datasets and conducted Human vs. AI text detection using state-of-the-art (SOTA) models. While SOTA methods, such as, character n-gram or Transformer-based model, exhibit similar AA {\&} AV performance in human-spoken datasets compared to written ones, there is much room for improvement in AI-generated spoken text detection. The ${\sf HANSEN}$ benchmark is available at: https://huggingface.co/datasets/HANSEN-REPO/HANSEN

HANSEN: Human and AI Spoken Text Benchmark for Authorship Analysis

findings-emnlp.916

2310.16746

Poster

https://aclanthology.org/2023.findings-emnlp.917.bib

https://aclanthology.org/2023.findings-emnlp.917/

@inproceedings{xie-etal-2023-data, title = "Data Augmentation for Code Translation with Comparable Corpora and Multiple References", author = "Xie, Yiqing and Naik, Atharva and Fried, Daniel and Rose, Carolyn", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.917", doi = "10.18653/v1/2023.findings-emnlp.917", pages = "13725--13739", abstract = "One major challenge of translating code between programming languages is that parallel training data is often limited. To overcome this challenge, we present two data augmentation techniques, one that builds comparable corpora (i.e., code pairs with similar functionality), and another that augments existing parallel data with multiple reference translations. Specifically, we build and analyze multiple types of comparable corpora, including programs generated from natural language documentation using a code generation model. Furthermore, to reduce overfitting to a single reference translation, we automatically generate additional translation references for available parallel data and filter the translations by unit tests, which increases variation in target translations. Experiments show that our data augmentation techniques significantly improve CodeT5 for translation between Java, Python, and C++ by an average of 7.5{\%} Computational Accuracy (CA@1), which verifies the correctness of translations by execution. The code is available at https://github.com/Veronicium/CMTrans.", }

One major challenge of translating code between programming languages is that parallel training data is often limited. To overcome this challenge, we present two data augmentation techniques, one that builds comparable corpora (i.e., code pairs with similar functionality), and another that augments existing parallel data with multiple reference translations. Specifically, we build and analyze multiple types of comparable corpora, including programs generated from natural language documentation using a code generation model. Furthermore, to reduce overfitting to a single reference translation, we automatically generate additional translation references for available parallel data and filter the translations by unit tests, which increases variation in target translations. Experiments show that our data augmentation techniques significantly improve CodeT5 for translation between Java, Python, and C++ by an average of 7.5{\%} Computational Accuracy (CA@1), which verifies the correctness of translations by execution. The code is available at https://github.com/Veronicium/CMTrans.

Data Augmentation for Code Translation with Comparable Corpora and Multiple References

findings-emnlp.917

2311.00317

Poster

https://aclanthology.org/2023.findings-emnlp.918.bib

https://aclanthology.org/2023.findings-emnlp.918/

@inproceedings{han-gardent-2023-multilingual, title = "Multilingual Generation and Answering of Questions from Texts and Knowledge Graphs", author = "Han, Kelvin and Gardent, Claire", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.918", doi = "10.18653/v1/2023.findings-emnlp.918", pages = "13740--13756", abstract = "The ability to bridge Question Generation (QG) and Question Answering (QA) across structured and unstructured modalities has the potential for aiding different NLP applications. One key application is in QA-based methods that have recently been shown to be useful for automatically evaluating Natural Language (NL) texts generated from Knowledge Graphs (KG). While methods have been proposed for QG-QA across these modalities, these efforts have been in English only; in this work, we bring multilinguality (Brazilian Portuguese and Russian) to multimodal (KG and NL) QG-QA. Using synthetic data generation and machine translation to produce QG-QA data that is aligned between graph and text, we are able to train multimodal, multi-task models that can perform multimodal QG and QA in Portuguese and Russian. We show that our approach outperforms a baseline which is derived from previous work on English and adapted to handle these two languages.", }

The ability to bridge Question Generation (QG) and Question Answering (QA) across structured and unstructured modalities has the potential for aiding different NLP applications. One key application is in QA-based methods that have recently been shown to be useful for automatically evaluating Natural Language (NL) texts generated from Knowledge Graphs (KG). While methods have been proposed for QG-QA across these modalities, these efforts have been in English only; in this work, we bring multilinguality (Brazilian Portuguese and Russian) to multimodal (KG and NL) QG-QA. Using synthetic data generation and machine translation to produce QG-QA data that is aligned between graph and text, we are able to train multimodal, multi-task models that can perform multimodal QG and QA in Portuguese and Russian. We show that our approach outperforms a baseline which is derived from previous work on English and adapted to handle these two languages.

Multilingual Generation and Answering of Questions from Texts and Knowledge Graphs

findings-emnlp.918

Poster

https://aclanthology.org/2023.findings-emnlp.919.bib

https://aclanthology.org/2023.findings-emnlp.919/

@inproceedings{wang-etal-2023-infodiffusion, title = "{I}nfo{D}iffusion: Information Entropy Aware Diffusion Process for Non-Autoregressive Text Generation", author = "Wang, Renzhi and Li, Jing and Li, Piji", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2023", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.findings-emnlp.919", doi = "10.18653/v1/2023.findings-emnlp.919", pages = "13757--13770", abstract = "Diffusion models have garnered considerable interest in the field of text generation. Several studies have explored text diffusion models with different structures and applied them to various tasks, including named entity recognition and summarization. However, there exists a notable disparity between the {``}easy-first{''} text generation process of current diffusion models and the {``}keyword-first{''} natural text generation process of humans, which has received limited attention. To bridge this gap, we propose InfoDiffusion, a non-autoregressive text diffusion model. Our approach introduces a {``}keyinfo-first{''} generation strategy and incorporates a noise schedule based on the amount of text information. In addition, InfoDiffusion combines self-conditioning with a newly proposed partially noising model structure. Experimental results show that InfoDiffusion outperforms the baseline model in terms of generation quality and diversity, as well as exhibiting higher sampling efficiency.", }

Diffusion models have garnered considerable interest in the field of text generation. Several studies have explored text diffusion models with different structures and applied them to various tasks, including named entity recognition and summarization. However, there exists a notable disparity between the {``}easy-first{''} text generation process of current diffusion models and the {``}keyword-first{''} natural text generation process of humans, which has received limited attention. To bridge this gap, we propose InfoDiffusion, a non-autoregressive text diffusion model. Our approach introduces a {``}keyinfo-first{''} generation strategy and incorporates a noise schedule based on the amount of text information. In addition, InfoDiffusion combines self-conditioning with a newly proposed partially noising model structure. Experimental results show that InfoDiffusion outperforms the baseline model in terms of generation quality and diversity, as well as exhibiting higher sampling efficiency.

InfoDiffusion: Information Entropy Aware Diffusion Process for Non-Autoregressive Text Generation

findings-emnlp.919

2310.11976

https://huggingface.co/papers/2310.11976

Poster