Datasets:

IAMJB

/

EMNLP2023-papers

like 0

https://aclanthology.org/2023.emnlp-main.601.bib

https://aclanthology.org/2023.emnlp-main.601/

@inproceedings{park-etal-2023-dive, title = "{DIVE}: Towards Descriptive and Diverse Visual Commonsense Generation", author = "Park, Jun-Hyung and Park, Hyuntae and Kang, Youjin and Jeon, Eojin and Lee, SangKeun", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.601", doi = "10.18653/v1/2023.emnlp-main.601", pages = "9677--9695", abstract = "Towards human-level visual understanding, visual commonsense generation has been introduced to generate commonsense inferences beyond images. However, current research on visual commonsense generation has overlooked an important human cognitive ability: generating descriptive and diverse inferences. In this work, we propose a novel visual commonsense generation framework, called DIVE, which aims to improve the descriptiveness and diversity of generated inferences. DIVE involves two methods, generic inference filtering and contrastive retrieval learning, which address the limitations of existing visual commonsense resources and training objectives. Experimental results verify that DIVE outperforms state-of-the-art models for visual commonsense generation in terms of both descriptiveness and diversity, while showing a superior quality in generating unique and novel inferences. Notably, DIVE achieves human-level descriptiveness and diversity on Visual Commonsense Graphs. Furthermore, human evaluations confirm that DIVE aligns closely with human judgments on descriptiveness and diversity.", }

Towards human-level visual understanding, visual commonsense generation has been introduced to generate commonsense inferences beyond images. However, current research on visual commonsense generation has overlooked an important human cognitive ability: generating descriptive and diverse inferences. In this work, we propose a novel visual commonsense generation framework, called DIVE, which aims to improve the descriptiveness and diversity of generated inferences. DIVE involves two methods, generic inference filtering and contrastive retrieval learning, which address the limitations of existing visual commonsense resources and training objectives. Experimental results verify that DIVE outperforms state-of-the-art models for visual commonsense generation in terms of both descriptiveness and diversity, while showing a superior quality in generating unique and novel inferences. Notably, DIVE achieves human-level descriptiveness and diversity on Visual Commonsense Graphs. Furthermore, human evaluations confirm that DIVE aligns closely with human judgments on descriptiveness and diversity.

DIVE: Towards Descriptive and Diverse Visual Commonsense Generation

emnlp-main.601

Poster

https://aclanthology.org/2023.emnlp-main.602.bib

https://aclanthology.org/2023.emnlp-main.602/

@inproceedings{nguyen-etal-2023-towards, title = "Towards Conceptualization of {``}Fair Explanation{''}: Disparate Impacts of anti-{A}sian Hate Speech Explanations on Content Moderators", author = "Nguyen, Tin and Xu, Jiannan and Roy, Aayushi and Daum{\'e} III, Hal and Carpuat, Marine", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.602", doi = "10.18653/v1/2023.emnlp-main.602", pages = "9696--9717", abstract = "Recent research at the intersection of AI explainability and fairness has focused on how explanations can improve human-plus-AI task performance as assessed by fairness measures. We propose to characterize what constitutes an explanation that is itself {``}fair{''} {--} an explanation that does not adversely impact specific populations. We formulate a novel evaluation method of {``}fair explanations{''} using not just accuracy and label time, but also psychological impact of explanations on different user groups across many metrics (mental discomfort, stereotype activation, and perceived workload). We apply this method in the context of content moderation of potential hate speech, and its differential impact on Asian vs. non-Asian proxy moderators, across explanation approaches (saliency map and counterfactual explanation). We find that saliency maps generally perform better and show less evidence of disparate impact (group) and individual unfairness than counterfactual explanations. Content warning: This paper contains examples of hate speech and racially discriminatory language. The authors do not support such content. Please consider your risk of discomfort carefully before continuing reading!", }

Recent research at the intersection of AI explainability and fairness has focused on how explanations can improve human-plus-AI task performance as assessed by fairness measures. We propose to characterize what constitutes an explanation that is itself {``}fair{''} {--} an explanation that does not adversely impact specific populations. We formulate a novel evaluation method of {``}fair explanations{''} using not just accuracy and label time, but also psychological impact of explanations on different user groups across many metrics (mental discomfort, stereotype activation, and perceived workload). We apply this method in the context of content moderation of potential hate speech, and its differential impact on Asian vs. non-Asian proxy moderators, across explanation approaches (saliency map and counterfactual explanation). We find that saliency maps generally perform better and show less evidence of disparate impact (group) and individual unfairness than counterfactual explanations. Content warning: This paper contains examples of hate speech and racially discriminatory language. The authors do not support such content. Please consider your risk of discomfort carefully before continuing reading!

Towards Conceptualization of “Fair Explanation”: Disparate Impacts of anti-Asian Hate Speech Explanations on Content Moderators

emnlp-main.602

Poster

https://aclanthology.org/2023.emnlp-main.603.bib

https://aclanthology.org/2023.emnlp-main.603/

@inproceedings{han-etal-2023-bridging, title = "Bridging Background Knowledge Gaps in Translation with Automatic Explicitation", author = "Han, HyoJung and Boyd-Graber, Jordan and Carpuat, Marine", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.603", doi = "10.18653/v1/2023.emnlp-main.603", pages = "9718--9735", abstract = "Translations help people understand content written in another language. However, even correct literal translations do not fulfill that goal when people lack the necessary background to understand them. Professional translators incorporate explicitations to explain the missing context by considering cultural differences between source and target audiences. Despite its potential to help users, NLP research on explicitation is limited because of the dearth of adequate evaluation methods. This work introduces techniques for automatically generating explicitations, motivated by WikiExpl: a dataset that we collect from Wikipedia and annotate with human translators. The resulting explicitations are useful as they help answer questions more accurately in a multilingual question answering framework.", }

Translations help people understand content written in another language. However, even correct literal translations do not fulfill that goal when people lack the necessary background to understand them. Professional translators incorporate explicitations to explain the missing context by considering cultural differences between source and target audiences. Despite its potential to help users, NLP research on explicitation is limited because of the dearth of adequate evaluation methods. This work introduces techniques for automatically generating explicitations, motivated by WikiExpl: a dataset that we collect from Wikipedia and annotate with human translators. The resulting explicitations are useful as they help answer questions more accurately in a multilingual question answering framework.

Bridging Background Knowledge Gaps in Translation with Automatic Explicitation

emnlp-main.603

2312.01308

Poster

https://aclanthology.org/2023.emnlp-main.604.bib

https://aclanthology.org/2023.emnlp-main.604/

@inproceedings{zhang-etal-2023-quality, title = "A Quality-based Syntactic Template Retriever for Syntactically-Controlled Paraphrase Generation", author = "Zhang, Xue and Zhang, Songming and Liang, Yunlong and Chen, Yufeng and Liu, Jian and Han, Wenjuan and Xu, Jinan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.604", doi = "10.18653/v1/2023.emnlp-main.604", pages = "9736--9748", abstract = "Existing syntactically-controlled paraphrase generation (SPG) models perform promisingly with human-annotated or well-chosen syntactic templates. However, the difficulty of obtaining such templates actually hinders the practical application of SPG models. For one thing, the prohibitive cost makes it unfeasible to manually design decent templates for every source sentence. For another, the templates automatically retrieved by current heuristic methods are usually unreliable for SPG models to generate qualified paraphrases. To escape this dilemma, we propose a novel Quality-based Syntactic Template Retriever (QSTR) to retrieve templates based on the quality of the to-be-generated paraphrases. Furthermore, for situations requiring multiple paraphrases for each source sentence, we design a Diverse Templates Search (DTS) algorithm, which can enhance the diversity between paraphrases without sacrificing quality. Experiments demonstrate that QSTR can significantly surpass existing retrieval methods in generating high-quality paraphrases and even perform comparably with human-annotated templates in terms of reference-free metrics. Additionally, human evaluation and the performance on downstream tasks using our generated paraphrases for data augmentation showcase the potential of our QSTR and DTS algorithm in practical scenarios.", }

Existing syntactically-controlled paraphrase generation (SPG) models perform promisingly with human-annotated or well-chosen syntactic templates. However, the difficulty of obtaining such templates actually hinders the practical application of SPG models. For one thing, the prohibitive cost makes it unfeasible to manually design decent templates for every source sentence. For another, the templates automatically retrieved by current heuristic methods are usually unreliable for SPG models to generate qualified paraphrases. To escape this dilemma, we propose a novel Quality-based Syntactic Template Retriever (QSTR) to retrieve templates based on the quality of the to-be-generated paraphrases. Furthermore, for situations requiring multiple paraphrases for each source sentence, we design a Diverse Templates Search (DTS) algorithm, which can enhance the diversity between paraphrases without sacrificing quality. Experiments demonstrate that QSTR can significantly surpass existing retrieval methods in generating high-quality paraphrases and even perform comparably with human-annotated templates in terms of reference-free metrics. Additionally, human evaluation and the performance on downstream tasks using our generated paraphrases for data augmentation showcase the potential of our QSTR and DTS algorithm in practical scenarios.

A Quality-based Syntactic Template Retriever for Syntactically-Controlled Paraphrase Generation

emnlp-main.604

2310.13262

Poster

https://aclanthology.org/2023.emnlp-main.605.bib

https://aclanthology.org/2023.emnlp-main.605/

@inproceedings{wu-monz-2023-beyond, title = "Beyond Shared Vocabulary: Increasing Representational Word Similarities across Languages for Multilingual Machine Translation", author = "Wu, Di and Monz, Christof", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.605", doi = "10.18653/v1/2023.emnlp-main.605", pages = "9749--9764", abstract = "Using a shared vocabulary is common practice in Multilingual Neural Machine Translation (MNMT). In addition to its simple design, shared tokens play an important role in positive knowledge transfer, which manifests naturally when the shared tokens refer to similar meanings across languages. However, when words overlap is small, e.g., using different writing systems, transfer is inhibited. In this paper, we propose a re-parameterized method for building embeddings to alleviate this problem. More specifically, we define word-level information transfer pathways via word equivalence classes and rely on graph networks to fuse word embeddings across languages. Our experiments demonstrate the advantages of our approach: 1) the semantics of embeddings are better aligned across languages, 2) our method achieves evident BLEU improvements on high- and low-resource MNMT, and 3) only less than 1.0{\%} additional trainable parameters are required with a limited increase in computational costs, while the inference time is identical to baselines.", }

Using a shared vocabulary is common practice in Multilingual Neural Machine Translation (MNMT). In addition to its simple design, shared tokens play an important role in positive knowledge transfer, which manifests naturally when the shared tokens refer to similar meanings across languages. However, when words overlap is small, e.g., using different writing systems, transfer is inhibited. In this paper, we propose a re-parameterized method for building embeddings to alleviate this problem. More specifically, we define word-level information transfer pathways via word equivalence classes and rely on graph networks to fuse word embeddings across languages. Our experiments demonstrate the advantages of our approach: 1) the semantics of embeddings are better aligned across languages, 2) our method achieves evident BLEU improvements on high- and low-resource MNMT, and 3) only less than 1.0{\%} additional trainable parameters are required with a limited increase in computational costs, while the inference time is identical to baselines.

Beyond Shared Vocabulary: Increasing Representational Word Similarities across Languages for Multilingual Machine Translation

emnlp-main.605

2305.14189

Poster

https://aclanthology.org/2023.emnlp-main.606.bib

https://aclanthology.org/2023.emnlp-main.606/

@inproceedings{wolf-etal-2023-quantifying, title = "Quantifying the redundancy between prosody and text", author = "Wolf, Lukas and Pimentel, Tiago and Fedorenko, Evelina and Cotterell, Ryan and Warstadt, Alex and Wilcox, Ethan and Regev, Tamar", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.606", doi = "10.18653/v1/2023.emnlp-main.606", pages = "9765--9784", abstract = "Prosody{---}the suprasegmental component of speech, including pitch, loudness, and tempo{---}carries critical aspects of meaning. However, the relationship between the information conveyed by prosody vs. by the words themselves remains poorly understood. We use large language models (LLMs) to estimate how much information is redundant between prosody and the words themselves. Using a large spoken corpus of English audiobooks, we extract prosodic features aligned to individual words and test how well they can be predicted from LLM embeddings, compared to non-contextual word embeddings. We find a high degree of redundancy between the information carried by the words and prosodic information across several prosodic features, including intensity, duration, pauses, and pitch contours. Furthermore, a word{'}s prosodic information is redundant with both the word itself and the context preceding as well as following it. Still, we observe that prosodic features can not be fully predicted from text, suggesting that prosody carries information above and beyond the words. Along with this paper, we release a general-purpose data processing pipeline for quantifying the relationship between linguistic information and extra-linguistic features.", }

Prosody{---}the suprasegmental component of speech, including pitch, loudness, and tempo{---}carries critical aspects of meaning. However, the relationship between the information conveyed by prosody vs. by the words themselves remains poorly understood. We use large language models (LLMs) to estimate how much information is redundant between prosody and the words themselves. Using a large spoken corpus of English audiobooks, we extract prosodic features aligned to individual words and test how well they can be predicted from LLM embeddings, compared to non-contextual word embeddings. We find a high degree of redundancy between the information carried by the words and prosodic information across several prosodic features, including intensity, duration, pauses, and pitch contours. Furthermore, a word{'}s prosodic information is redundant with both the word itself and the context preceding as well as following it. Still, we observe that prosodic features can not be fully predicted from text, suggesting that prosody carries information above and beyond the words. Along with this paper, we release a general-purpose data processing pipeline for quantifying the relationship between linguistic information and extra-linguistic features.

Quantifying the redundancy between prosody and text

emnlp-main.606

2311.17233

Oral

https://aclanthology.org/2023.emnlp-main.607.bib

https://aclanthology.org/2023.emnlp-main.607/

@inproceedings{ismayilzada-etal-2023-crow, title = "{CR}o{W}: Benchmarking Commonsense Reasoning in Real-World Tasks", author = "Ismayilzada, Mete and Paul, Debjit and Montariol, Syrielle and Geva, Mor and Bosselut, Antoine", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.607", doi = "10.18653/v1/2023.emnlp-main.607", pages = "9785--9821", abstract = "Recent efforts in natural language processing (NLP) commonsense reasoning research have yielded a considerable number of new datasets and benchmarks. However, most of these datasets formulate commonsense reasoning challenges in artificial scenarios that are not reflective of the tasks which real-world NLP systems are designed to solve. In this work, we present CRoW, a manually-curated, multi-task benchmark that evaluates the ability of models to apply commonsense reasoning in the context of six real-world NLP tasks. CRoW is constructed using a multi-stage data collection pipeline that rewrites examples from existing datasets using commonsense-violating perturbations. We use CRoW to study how NLP systems perform across different dimensions of commonsense knowledge, such as physical, temporal, and social reasoning. We find a significant performance gap when NLP systems are evaluated on CRoW compared to humans, showcasing that commonsense reasoning is far from being solved in real-world task settings. We make our dataset and leaderboard available to the research community.", }

Recent efforts in natural language processing (NLP) commonsense reasoning research have yielded a considerable number of new datasets and benchmarks. However, most of these datasets formulate commonsense reasoning challenges in artificial scenarios that are not reflective of the tasks which real-world NLP systems are designed to solve. In this work, we present CRoW, a manually-curated, multi-task benchmark that evaluates the ability of models to apply commonsense reasoning in the context of six real-world NLP tasks. CRoW is constructed using a multi-stage data collection pipeline that rewrites examples from existing datasets using commonsense-violating perturbations. We use CRoW to study how NLP systems perform across different dimensions of commonsense knowledge, such as physical, temporal, and social reasoning. We find a significant performance gap when NLP systems are evaluated on CRoW compared to humans, showcasing that commonsense reasoning is far from being solved in real-world task settings. We make our dataset and leaderboard available to the research community.

CRoW: Benchmarking Commonsense Reasoning in Real-World Tasks

emnlp-main.607

Oral

https://aclanthology.org/2023.emnlp-main.608.bib

https://aclanthology.org/2023.emnlp-main.608/

@inproceedings{bhattacharyya-etal-2023-video, title = "A Video Is Worth 4096 Tokens: Verbalize Videos To Understand Them In Zero Shot", author = "Bhattacharyya, Aanisha and Singla, Yaman K and Krishnamurthy, Balaji and Shah, Rajiv Ratn and Chen, Changyou", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.608", doi = "10.18653/v1/2023.emnlp-main.608", pages = "9822--9839", abstract = "Multimedia content, such as advertisements and story videos, exhibit a rich blend of creativity and multiple modalities. They incorporate elements like text, visuals, audio, and storytelling techniques, employing devices like emotions, symbolism, and slogans to convey meaning. There is a dearth of large annotated training datasets in the multimedia domain hindering the development of supervised learning models with satisfactory performance for real-world applications. On the other hand, the rise of large language models (LLMs) has witnessed remarkable zero-shot performance in various natural language processing (NLP) tasks, such as emotion classification, question answering, and topic classification. To leverage such advanced techniques to bridge this performance gap in multimedia understanding, we propose verbalizing long videos to generate their descriptions in natural language, followed by performing video-understanding tasks on the generated story as opposed to the original video. Through extensive experiments on fifteen video-understanding tasks, we demonstrate that our method, despite being zero-shot, achieves significantly better results than supervised baselines for video understanding. Furthermore, to alleviate a lack of story understanding benchmarks, we publicly release the first dataset on a crucial task in computational social science on persuasion strategy identification.", }

Multimedia content, such as advertisements and story videos, exhibit a rich blend of creativity and multiple modalities. They incorporate elements like text, visuals, audio, and storytelling techniques, employing devices like emotions, symbolism, and slogans to convey meaning. There is a dearth of large annotated training datasets in the multimedia domain hindering the development of supervised learning models with satisfactory performance for real-world applications. On the other hand, the rise of large language models (LLMs) has witnessed remarkable zero-shot performance in various natural language processing (NLP) tasks, such as emotion classification, question answering, and topic classification. To leverage such advanced techniques to bridge this performance gap in multimedia understanding, we propose verbalizing long videos to generate their descriptions in natural language, followed by performing video-understanding tasks on the generated story as opposed to the original video. Through extensive experiments on fifteen video-understanding tasks, we demonstrate that our method, despite being zero-shot, achieves significantly better results than supervised baselines for video understanding. Furthermore, to alleviate a lack of story understanding benchmarks, we publicly release the first dataset on a crucial task in computational social science on persuasion strategy identification.

A Video Is Worth 4096 Tokens: Verbalize Videos To Understand Them In Zero Shot

emnlp-main.608

2305.09758

https://huggingface.co/papers/2305.09758

Oral

https://aclanthology.org/2023.emnlp-main.609.bib

https://aclanthology.org/2023.emnlp-main.609/

@inproceedings{wang-etal-2023-label, title = "Label Words are Anchors: An Information Flow Perspective for Understanding In-Context Learning", author = "Wang, Lean and Li, Lei and Dai, Damai and Chen, Deli and Zhou, Hao and Meng, Fandong and Zhou, Jie and Sun, Xu", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.609", doi = "10.18653/v1/2023.emnlp-main.609", pages = "9840--9855", abstract = "In-context learning (ICL) emerges as a promising capability of large language models (LLMs) by providing them with demonstration examples to perform diverse tasks. However, the underlying mechanism of how LLMs learn from the provided context remains under-explored. In this paper, we investigate the working mechanism of ICL through an information flow lens. Our findings reveal that label words in the demonstration examples function as anchors: (1) semantic information aggregates into label word representations during the shallow computation layers{'} processing; (2) the consolidated information in label words serves as a reference for LLMs{'} final predictions. Based on these insights, we introduce an anchor re-weighting method to improve ICL performance, a demonstration compression technique to expedite inference, and an analysis framework for diagnosing ICL errors in GPT2-XL. The promising applications of our findings again validate the uncovered ICL working mechanism and pave the way for future studies.", }

In-context learning (ICL) emerges as a promising capability of large language models (LLMs) by providing them with demonstration examples to perform diverse tasks. However, the underlying mechanism of how LLMs learn from the provided context remains under-explored. In this paper, we investigate the working mechanism of ICL through an information flow lens. Our findings reveal that label words in the demonstration examples function as anchors: (1) semantic information aggregates into label word representations during the shallow computation layers{'} processing; (2) the consolidated information in label words serves as a reference for LLMs{'} final predictions. Based on these insights, we introduce an anchor re-weighting method to improve ICL performance, a demonstration compression technique to expedite inference, and an analysis framework for diagnosing ICL errors in GPT2-XL. The promising applications of our findings again validate the uncovered ICL working mechanism and pave the way for future studies.

Label Words are Anchors: An Information Flow Perspective for Understanding In-Context Learning

emnlp-main.609

2305.14160

https://huggingface.co/papers/2305.14160

Poster

https://aclanthology.org/2023.emnlp-main.610.bib

https://aclanthology.org/2023.emnlp-main.610/

@inproceedings{parashar-etal-2023-prompting, title = "Prompting Scientific Names for Zero-Shot Species Recognition", author = "Parashar, Shubham and Lin, Zhiqiu and Li, Yanan and Kong, Shu", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.610", doi = "10.18653/v1/2023.emnlp-main.610", pages = "9856--9861", abstract = "Trained on web-scale image-text pairs, Vision-Language Models (VLMs) such as CLIP can recognize images of common objects in a zero-shot fashion. However, it is underexplored how to use CLIP for zero-shot recognition of highly specialized concepts, e.g., species of birds, plants, and animals, for which their scientific names are written in Latin or Greek. Indeed, CLIP performs poorly for zero-shot species recognition with prompts that use scientific names, e.g., {``}a photo of Lepus Timidus{''} (which is a scientific name in Latin). This is because these names are usually not included in CLIP{'}s training set. To improve performance, we explore using large-language models (LLMs) to generate descriptions (e.g., of species color and shape) and additionally use them in prompts. However, this method improves only marginally. Instead, we are motivated to translate scientific names (e.g., Lepus Timidus) to common English names (e.g., mountain hare) and use such in the prompts. We find that common names are more likely to be included in CLIP{'}s training set, and prompting them achieves 2{\textasciitilde}5 times higher accuracy on benchmarking datasets of fine-grained species recognition.", }

Trained on web-scale image-text pairs, Vision-Language Models (VLMs) such as CLIP can recognize images of common objects in a zero-shot fashion. However, it is underexplored how to use CLIP for zero-shot recognition of highly specialized concepts, e.g., species of birds, plants, and animals, for which their scientific names are written in Latin or Greek. Indeed, CLIP performs poorly for zero-shot species recognition with prompts that use scientific names, e.g., {``}a photo of Lepus Timidus{''} (which is a scientific name in Latin). This is because these names are usually not included in CLIP{'}s training set. To improve performance, we explore using large-language models (LLMs) to generate descriptions (e.g., of species color and shape) and additionally use them in prompts. However, this method improves only marginally. Instead, we are motivated to translate scientific names (e.g., Lepus Timidus) to common English names (e.g., mountain hare) and use such in the prompts. We find that common names are more likely to be included in CLIP{'}s training set, and prompting them achieves 2{\textasciitilde}5 times higher accuracy on benchmarking datasets of fine-grained species recognition.

Prompting Scientific Names for Zero-Shot Species Recognition

emnlp-main.610

2310.09929

Poster

https://aclanthology.org/2023.emnlp-main.611.bib

https://aclanthology.org/2023.emnlp-main.611/

@inproceedings{perlitz-etal-2023-active, title = "Active Learning for Natural Language Generation", author = "Perlitz, Yotam and Gera, Ariel and Shmueli-Scheuer, Michal and Sheinwald, Dafna and Slonim, Noam and Ein-Dor, Liat", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.611", doi = "10.18653/v1/2023.emnlp-main.611", pages = "9862--9877", abstract = "The field of Natural Language Generation (NLG) suffers from a severe shortage of labeled data due to the extremely expensive and time-consuming process involved in manual annotation. A natural approach for coping with this problem is active learning (AL), a well-known machine learning technique for improving annotation efficiency by selectively choosing the most informative examples to label. However, while AL has been well-researched in the context of text classification, its application to NLG remains largely unexplored. In this paper, we present a first systematic study of active learning for NLG, considering a diverse set of tasks and multiple leading selection strategies, and harnessing a strong instruction-tuned model. Our results indicate that the performance of existing AL strategies is inconsistent, surpassing the baseline of random example selection in some cases but not in others. We highlight some notable differences between the classification and generation scenarios, and analyze the selection behaviors of existing AL strategies. Our findings motivate exploring novel approaches for applying AL to generation tasks.", }

The field of Natural Language Generation (NLG) suffers from a severe shortage of labeled data due to the extremely expensive and time-consuming process involved in manual annotation. A natural approach for coping with this problem is active learning (AL), a well-known machine learning technique for improving annotation efficiency by selectively choosing the most informative examples to label. However, while AL has been well-researched in the context of text classification, its application to NLG remains largely unexplored. In this paper, we present a first systematic study of active learning for NLG, considering a diverse set of tasks and multiple leading selection strategies, and harnessing a strong instruction-tuned model. Our results indicate that the performance of existing AL strategies is inconsistent, surpassing the baseline of random example selection in some cases but not in others. We highlight some notable differences between the classification and generation scenarios, and analyze the selection behaviors of existing AL strategies. Our findings motivate exploring novel approaches for applying AL to generation tasks.

Active Learning for Natural Language Generation

emnlp-main.611

Oral

https://aclanthology.org/2023.emnlp-main.612.bib

https://aclanthology.org/2023.emnlp-main.612/

@inproceedings{wen-etal-2023-re3dial, title = "{R}e$^3${D}ial: Retrieve, Reorganize and Rescale Conversations for Long-Turn Open-Domain Dialogue Pre-training", author = "Wen, Jiaxin and Zhou, Hao and Guan, Jian and Zhou, Jie and Huang, Minlie", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.612", doi = "10.18653/v1/2023.emnlp-main.612", pages = "9878--9894", abstract = "Pre-training on large-scale open-domain dialogue data can substantially improve the performance of dialogue models. However, the pre-trained dialogue model{'}s ability to utilize long-range context is limited due to the scarcity of long-turn dialogue sessions. Most dialogues in existing pre-training corpora contain fewer than three turns of dialogue. To alleviate this issue, we propose the Retrieve, Reorganize and Rescale framework (Re$^3$Dial), which can automatically construct billion-scale long-turn dialogues by reorganizing existing short-turn ones. Given a short-turn session, Re$^3$Dial first employs a session retriever to retrieve coherent consecutive sessions. To this end, we train the retriever to capture semantic and discourse relations within multi-turn dialogues through contrastive training. Next, Re$^3$Dial samples a session from retrieved results following a diversity sampling strategy, which is designed to penalize repetitive or generic sessions. A longer session is then derived by concatenating the original session and the sampled session. By repeating the above process, Re$^3$Dial can yield a coherent long-turn dialogue. Extensive experiments on multiple multi-turn dialogue benchmarks demonstrate that Re$^3$Dial significantly improves the dialogue model{'}s ability to utilize long-range context and thus generate more sensible and informative responses. Finally, we build a toolkit for efficiently rescaling conversations with Re$^3$Dial, which enables us to construct a corpus containing 1B Chinese dialogue sessions with 11.3 turns on average (5X longer than the original corpus). We will release our retriever model, toolkit, and data for public use.", }

Pre-training on large-scale open-domain dialogue data can substantially improve the performance of dialogue models. However, the pre-trained dialogue model{'}s ability to utilize long-range context is limited due to the scarcity of long-turn dialogue sessions. Most dialogues in existing pre-training corpora contain fewer than three turns of dialogue. To alleviate this issue, we propose the Retrieve, Reorganize and Rescale framework (Re$^3$Dial), which can automatically construct billion-scale long-turn dialogues by reorganizing existing short-turn ones. Given a short-turn session, Re$^3$Dial first employs a session retriever to retrieve coherent consecutive sessions. To this end, we train the retriever to capture semantic and discourse relations within multi-turn dialogues through contrastive training. Next, Re$^3$Dial samples a session from retrieved results following a diversity sampling strategy, which is designed to penalize repetitive or generic sessions. A longer session is then derived by concatenating the original session and the sampled session. By repeating the above process, Re$^3$Dial can yield a coherent long-turn dialogue. Extensive experiments on multiple multi-turn dialogue benchmarks demonstrate that Re$^3$Dial significantly improves the dialogue model{'}s ability to utilize long-range context and thus generate more sensible and informative responses. Finally, we build a toolkit for efficiently rescaling conversations with Re$^3$Dial, which enables us to construct a corpus containing 1B Chinese dialogue sessions with 11.3 turns on average (5X longer than the original corpus). We will release our retriever model, toolkit, and data for public use.

Re^3Dial: Retrieve, Reorganize and Rescale Conversations for Long-Turn Open-Domain Dialogue Pre-training

emnlp-main.612

Poster

https://aclanthology.org/2023.emnlp-main.613.bib

https://aclanthology.org/2023.emnlp-main.613/

@inproceedings{shen-etal-2023-multiturncleanup, title = "{M}ulti{T}urn{C}leanup: A Benchmark for Multi-Turn Spoken Conversational Transcript Cleanup", author = "Shen, Hua and Zayats, Vicky and Rocholl, Johann and Walker, Daniel and Padfield, Dirk", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.613", doi = "10.18653/v1/2023.emnlp-main.613", pages = "9895--9903", abstract = "Current disfluency detection models focus on individual utterances each from a single speaker. However, numerous discontinuity phenomena in spoken conversational transcripts occur across multiple turns, which can not be identified by disfluency detection models. This study addresses these phenomena by proposing an innovative Multi-Turn Cleanup task for spoken conversational transcripts and collecting a new dataset, MultiTurnCleanup. We design a data labeling schema to collect the high-quality dataset and provide extensive data analysis. Furthermore, we leverage two modeling approaches for experimental evaluation as benchmarks for future research.", }

Current disfluency detection models focus on individual utterances each from a single speaker. However, numerous discontinuity phenomena in spoken conversational transcripts occur across multiple turns, which can not be identified by disfluency detection models. This study addresses these phenomena by proposing an innovative Multi-Turn Cleanup task for spoken conversational transcripts and collecting a new dataset, MultiTurnCleanup. We design a data labeling schema to collect the high-quality dataset and provide extensive data analysis. Furthermore, we leverage two modeling approaches for experimental evaluation as benchmarks for future research.

MultiTurnCleanup: A Benchmark for Multi-Turn Spoken Conversational Transcript Cleanup

emnlp-main.613

2305.12029

Oral

https://aclanthology.org/2023.emnlp-main.614.bib

https://aclanthology.org/2023.emnlp-main.614/

@inproceedings{ahia-etal-2023-languages, title = "Do All Languages Cost the Same? Tokenization in the Era of Commercial Language Models", author = "Ahia, Orevaoghene and Kumar, Sachin and Gonen, Hila and Kasai, Jungo and Mortensen, David and Smith, Noah and Tsvetkov, Yulia", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.614", doi = "10.18653/v1/2023.emnlp-main.614", pages = "9904--9923", abstract = "Language models have graduated from being research prototypes to commercialized products offered as web APIs, and recent works have highlighted the multilingual capabilities of these products. The API vendors charge their users based on usage, more specifically on the number of {``}tokens{''} processed or generated by the underlying language models. What constitutes a token, however, is training data and model dependent with a large variance in the number of tokens required to convey the same information in different languages. In this work, we analyze the effect of this non-uniformity on the fairness of an API{'}s pricing policy across languages. We conduct a systematic analysis of the cost and utility of OpenAI{'}s language model API on multilingual benchmarks in 22 typologically diverse languages. We show evidence that speakers of a large number of the supported languages are overcharged while obtaining poorer results. These speakers tend to also come from regions where the APIs are less affordable, to begin with. Through these analyses, we aim to increase transparency around language model APIs{'} pricing policies and encourage the vendors to make them more equitable.", }

Language models have graduated from being research prototypes to commercialized products offered as web APIs, and recent works have highlighted the multilingual capabilities of these products. The API vendors charge their users based on usage, more specifically on the number of {``}tokens{''} processed or generated by the underlying language models. What constitutes a token, however, is training data and model dependent with a large variance in the number of tokens required to convey the same information in different languages. In this work, we analyze the effect of this non-uniformity on the fairness of an API{'}s pricing policy across languages. We conduct a systematic analysis of the cost and utility of OpenAI{'}s language model API on multilingual benchmarks in 22 typologically diverse languages. We show evidence that speakers of a large number of the supported languages are overcharged while obtaining poorer results. These speakers tend to also come from regions where the APIs are less affordable, to begin with. Through these analyses, we aim to increase transparency around language model APIs{'} pricing policies and encourage the vendors to make them more equitable.

Do All Languages Cost the Same? Tokenization in the Era of Commercial Language Models

emnlp-main.614

2305.13707

Poster

https://aclanthology.org/2023.emnlp-main.615.bib

https://aclanthology.org/2023.emnlp-main.615/

@inproceedings{yu-etal-2023-characterizing, title = "Characterizing Mechanisms for Factual Recall in Language Models", author = "Yu, Qinan and Merullo, Jack and Pavlick, Ellie", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.615", doi = "10.18653/v1/2023.emnlp-main.615", pages = "9924--9959", abstract = "Language Models (LMs) often must integrate facts they memorized in pretraining with new information that appears in a given context. These two sources can disagree, causing competition within the model, and it is unclear how an LM will resolve the conflict. On a dataset that queries for knowledge of world capitals, we investigate both distributional and mechanistic determinants of LM behavior in such situations. Specifically, we measure the proportion of the time an LM will use a counterfactual prefix (e.g., {``}The capital of Poland is London{''}) to overwrite what it learned in pretraining ({``}Warsaw{''}). On Pythia and GPT2, the training frequency of both the query country ({''}Poland{''}) and the in-context city ({''}London{''}) highly affect the models{'} likelihood of using the counterfactual. We then use head attribution to identify individual attention heads that either promote the memorized answer or the in-context answer in the logits. By scaling up or down the value vector of these heads, we can control the likelihood of using the in-context answer on new data. This method can increase the rate of generating the in-context answer to 88{\%} of the time simply by scaling a single head at runtime. Our work contributes to a body of evidence showing that we can often localize model behaviors to specific components and provides a proof of concept for how future methods might control model behavior dynamically at runtime.", }

Language Models (LMs) often must integrate facts they memorized in pretraining with new information that appears in a given context. These two sources can disagree, causing competition within the model, and it is unclear how an LM will resolve the conflict. On a dataset that queries for knowledge of world capitals, we investigate both distributional and mechanistic determinants of LM behavior in such situations. Specifically, we measure the proportion of the time an LM will use a counterfactual prefix (e.g., {``}The capital of Poland is London{''}) to overwrite what it learned in pretraining ({``}Warsaw{''}). On Pythia and GPT2, the training frequency of both the query country ({''}Poland{''}) and the in-context city ({''}London{''}) highly affect the models{'} likelihood of using the counterfactual. We then use head attribution to identify individual attention heads that either promote the memorized answer or the in-context answer in the logits. By scaling up or down the value vector of these heads, we can control the likelihood of using the in-context answer on new data. This method can increase the rate of generating the in-context answer to 88{\%} of the time simply by scaling a single head at runtime. Our work contributes to a body of evidence showing that we can often localize model behaviors to specific components and provides a proof of concept for how future methods might control model behavior dynamically at runtime.

Characterizing Mechanisms for Factual Recall in Language Models

emnlp-main.615

Poster

https://aclanthology.org/2023.emnlp-main.616.bib

https://aclanthology.org/2023.emnlp-main.616/

@inproceedings{macko-etal-2023-multitude, title = "{MULTIT}u{DE}: Large-Scale Multilingual Machine-Generated Text Detection Benchmark", author = "Macko, Dominik and Moro, Robert and Uchendu, Adaku and Lucas, Jason and Yamashita, Michiharu and Pikuliak, Mat{\'u}{\v{s}} and Srba, Ivan and Le, Thai and Lee, Dongwon and Simko, Jakub and Bielikova, Maria", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.616", doi = "10.18653/v1/2023.emnlp-main.616", pages = "9960--9987", abstract = "There is a lack of research into capabilities of recent LLMs to generate convincing text in languages other than English and into performance of detectors of machine-generated text in multilingual settings. This is also reflected in the available benchmarks which lack authentic texts in languages other than English and predominantly cover older generators. To fill this gap, we introduce MULTITuDE, a novel benchmarking dataset for multilingual machine-generated text detection comprising of 74,081 authentic and machine-generated texts in 11 languages (ar, ca, cs, de, en, es, nl, pt, ru, uk, and zh) generated by 8 multilingual LLMs. Using this benchmark, we compare the performance of zero-shot (statistical and black-box) and fine-tuned detectors. Considering the multilinguality, we evaluate 1) how these detectors generalize to unseen languages (linguistically similar as well as dissimilar) and unseen LLMs and 2) whether the detectors improve their performance when trained on multiple languages.", }

There is a lack of research into capabilities of recent LLMs to generate convincing text in languages other than English and into performance of detectors of machine-generated text in multilingual settings. This is also reflected in the available benchmarks which lack authentic texts in languages other than English and predominantly cover older generators. To fill this gap, we introduce MULTITuDE, a novel benchmarking dataset for multilingual machine-generated text detection comprising of 74,081 authentic and machine-generated texts in 11 languages (ar, ca, cs, de, en, es, nl, pt, ru, uk, and zh) generated by 8 multilingual LLMs. Using this benchmark, we compare the performance of zero-shot (statistical and black-box) and fine-tuned detectors. Considering the multilinguality, we evaluate 1) how these detectors generalize to unseen languages (linguistically similar as well as dissimilar) and unseen LLMs and 2) whether the detectors improve their performance when trained on multiple languages.

MULTITuDE: Large-Scale Multilingual Machine-Generated Text Detection Benchmark

emnlp-main.616

Oral

https://aclanthology.org/2023.emnlp-main.617.bib

https://aclanthology.org/2023.emnlp-main.617/

@inproceedings{zhang-etal-2023-revisiting, title = "Revisiting Block-based Quantisation: What is Important for Sub-8-bit {LLM} Inference?", author = "Zhang, Cheng and Cheng, Jianyi and Shumailov, Ilia and Constantinides, George and Zhao, Yiren", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.617", doi = "10.18653/v1/2023.emnlp-main.617", pages = "9988--10006", abstract = "The inference of Large language models (LLMs) requires immense computation and memory resources. To curtail these costs, quantisation has emerged as a promising solution, but existing LLM quantisation mainly focuses on 8-bit. In this work, we explore the statistical and learning properties of the LLM layer and attribute the bottleneck of LLM quantisation to numerical scaling offsets. To address this, we adapt block quantisations for LLMs, a family of methods that share scaling factors across packed numbers. Block quantisations efficiently reduce the numerical scaling offsets solely from an arithmetic perspective, without additional treatments in the computational path. Our nearly-lossless quantised 6-bit LLMs achieve a $19\times$ higher arithmetic density and $5\times$ memory density than the float32 baseline, surpassing the prior art 8-bit quantisation by $2.5\times$ in arithmetic density and $1.2\times$ in memory density, without requiring any data calibration or re-training. We also share our insights into sub-8-bit LLM quantisation, including the mismatch between activation and weight distributions, optimal fine-tuning strategies, and a lower quantisation granularity inherent in the statistical properties of LLMs. The latter two tricks enable nearly-lossless 4-bit LLMs on downstream tasks. Our code is open-sourced.", }

The inference of Large language models (LLMs) requires immense computation and memory resources. To curtail these costs, quantisation has emerged as a promising solution, but existing LLM quantisation mainly focuses on 8-bit. In this work, we explore the statistical and learning properties of the LLM layer and attribute the bottleneck of LLM quantisation to numerical scaling offsets. To address this, we adapt block quantisations for LLMs, a family of methods that share scaling factors across packed numbers. Block quantisations efficiently reduce the numerical scaling offsets solely from an arithmetic perspective, without additional treatments in the computational path. Our nearly-lossless quantised 6-bit LLMs achieve a $19\times$ higher arithmetic density and $5\times$ memory density than the float32 baseline, surpassing the prior art 8-bit quantisation by $2.5\times$ in arithmetic density and $1.2\times$ in memory density, without requiring any data calibration or re-training. We also share our insights into sub-8-bit LLM quantisation, including the mismatch between activation and weight distributions, optimal fine-tuning strategies, and a lower quantisation granularity inherent in the statistical properties of LLMs. The latter two tricks enable nearly-lossless 4-bit LLMs on downstream tasks. Our code is open-sourced.

Revisiting Block-based Quantisation: What is Important for Sub-8-bit LLM Inference?

emnlp-main.617

2310.05079

https://huggingface.co/papers/2310.05079

Poster

https://aclanthology.org/2023.emnlp-main.618.bib

https://aclanthology.org/2023.emnlp-main.618/

@inproceedings{radhakrishnan-etal-2023-whispering, title = "Whispering {LL}a{MA}: A Cross-Modal Generative Error Correction Framework for Speech Recognition", author = "Radhakrishnan, Srijith and Yang, Chao-Han and Khan, Sumeer and Kumar, Rohit and Kiani, Narsis and Gomez-Cabrero, David and Tegn{\'e}r, Jesper", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.618", doi = "10.18653/v1/2023.emnlp-main.618", pages = "10007--10016", abstract = "We introduce a new cross-modal fusion technique designed for generative error correction in automatic speech recognition (ASR). Our methodology leverages both acoustic information and external linguistic representations to generate accurate speech transcription contexts. This marks a step towards a fresh paradigm in generative error correction within the realm of n-best hypotheses. Unlike the existing ranking-based rescoring methods, our approach adeptly uses distinct initialization techniques and parameter-efficient algorithms to boost ASR performance derived from pre-trained speech and text models. Through evaluation across diverse ASR datasets, we assess our fusion technique, demonstrating a 37.66{\%} improvement in word error rate (WER) relative performance compared to the n-best Oracle. To encourage future research, we have made our code and pre-trained models open source at [https://github.com/Srijith-rkr/Whispering-LLaMA](https://github.com/Srijith-rkr/Whispering-LLaMA)", }

We introduce a new cross-modal fusion technique designed for generative error correction in automatic speech recognition (ASR). Our methodology leverages both acoustic information and external linguistic representations to generate accurate speech transcription contexts. This marks a step towards a fresh paradigm in generative error correction within the realm of n-best hypotheses. Unlike the existing ranking-based rescoring methods, our approach adeptly uses distinct initialization techniques and parameter-efficient algorithms to boost ASR performance derived from pre-trained speech and text models. Through evaluation across diverse ASR datasets, we assess our fusion technique, demonstrating a 37.66{\%} improvement in word error rate (WER) relative performance compared to the n-best Oracle. To encourage future research, we have made our code and pre-trained models open source at [https://github.com/Srijith-rkr/Whispering-LLaMA](https://github.com/Srijith-rkr/Whispering-LLaMA)

Whispering LLaMA: A Cross-Modal Generative Error Correction Framework for Speech Recognition

emnlp-main.618

2310.06434

https://huggingface.co/papers/2310.06434

Oral

https://aclanthology.org/2023.emnlp-main.619.bib

https://aclanthology.org/2023.emnlp-main.619/

@inproceedings{kaneko-okazaki-2023-reducing, title = "Reducing Sequence Length by Predicting Edit Spans with Large Language Models", author = "Kaneko, Masahiro and Okazaki, Naoaki", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.619", doi = "10.18653/v1/2023.emnlp-main.619", pages = "10017--10029", abstract = "Large Language Models (LLMs) have demonstrated remarkable performance in various tasks and gained significant attention. LLMs are also used for local sequence transduction tasks, including grammatical error correction (GEC) and formality style transfer, where most tokens in a source text are kept unchanged. However, the models that generate all target tokens in such tasks have a tendency to simply copy the input text as is, without making needed changes, because the difference between input and output texts is minimal in the training data. This is also inefficient because the computational cost grows quadratically with the target sequence length with Transformer. This paper proposes predicting edit spans for the source text for local sequence transduction tasks. Representing an edit span with a position of the source text and corrected tokens, we can reduce the length of the target sequence and the computational cost for inference. We apply instruction tuning for LLMs on the supervision data of edit spans. Experiments show that the proposed method achieves comparable performance to the baseline in four tasks, paraphrasing, formality style transfer, GEC, and text simplification, despite reducing the length of the target text by as small as 21{\%}. Furthermore, we report that the task-specific fine-tuning with the proposed method achieved state-of-the-art performance in the four tasks.", }

Large Language Models (LLMs) have demonstrated remarkable performance in various tasks and gained significant attention. LLMs are also used for local sequence transduction tasks, including grammatical error correction (GEC) and formality style transfer, where most tokens in a source text are kept unchanged. However, the models that generate all target tokens in such tasks have a tendency to simply copy the input text as is, without making needed changes, because the difference between input and output texts is minimal in the training data. This is also inefficient because the computational cost grows quadratically with the target sequence length with Transformer. This paper proposes predicting edit spans for the source text for local sequence transduction tasks. Representing an edit span with a position of the source text and corrected tokens, we can reduce the length of the target sequence and the computational cost for inference. We apply instruction tuning for LLMs on the supervision data of edit spans. Experiments show that the proposed method achieves comparable performance to the baseline in four tasks, paraphrasing, formality style transfer, GEC, and text simplification, despite reducing the length of the target text by as small as 21{\%}. Furthermore, we report that the task-specific fine-tuning with the proposed method achieved state-of-the-art performance in the four tasks.

Reducing Sequence Length by Predicting Edit Spans with Large Language Models

emnlp-main.619

Poster

https://aclanthology.org/2023.emnlp-main.620.bib

https://aclanthology.org/2023.emnlp-main.620/

@inproceedings{jiao-etal-2023-instruct, title = "Instruct and Extract: Instruction Tuning for On-Demand Information Extraction", author = "Jiao, Yizhu and Zhong, Ming and Li, Sha and Zhao, Ruining and Ouyang, Siru and Ji, Heng and Han, Jiawei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.620", doi = "10.18653/v1/2023.emnlp-main.620", pages = "10030--10051", abstract = "Large language models with instruction-following capabilities open the door to a wider group of users. However, when it comes to information extraction {--} a classic task in natural language processing {--} most task-specific systems cannot align well with long-tail ad hoc extraction use cases for non-expert users. To address this, we propose a novel paradigm, termed On-Demand Information Extraction, to fulfill the personalized demands of real-world users. Our task aims to follow the instructions to extract the desired content from the associated text and present it in a structured tabular format. The table headers can either be user-specified or inferred contextually by the model. To facilitate research in this emerging area, we present a benchmark named InstructIE, inclusive of both automatically generated training data, as well as the human-annotated test set. Building on InstructIE, we further develop an On-Demand Information Extractor, ODIE. Comprehensive evaluations on our benchmark reveal that ODIE substantially outperforms the existing open-source models of similar size.", }

Large language models with instruction-following capabilities open the door to a wider group of users. However, when it comes to information extraction {--} a classic task in natural language processing {--} most task-specific systems cannot align well with long-tail ad hoc extraction use cases for non-expert users. To address this, we propose a novel paradigm, termed On-Demand Information Extraction, to fulfill the personalized demands of real-world users. Our task aims to follow the instructions to extract the desired content from the associated text and present it in a structured tabular format. The table headers can either be user-specified or inferred contextually by the model. To facilitate research in this emerging area, we present a benchmark named InstructIE, inclusive of both automatically generated training data, as well as the human-annotated test set. Building on InstructIE, we further develop an On-Demand Information Extractor, ODIE. Comprehensive evaluations on our benchmark reveal that ODIE substantially outperforms the existing open-source models of similar size.

Instruct and Extract: Instruction Tuning for On-Demand Information Extraction

emnlp-main.620

Oral

https://aclanthology.org/2023.emnlp-main.621.bib

https://aclanthology.org/2023.emnlp-main.621/

@inproceedings{wang-etal-2023-rethinking-evaluation, title = "Rethinking the Evaluation for Conversational Recommendation in the Era of Large Language Models", author = "Wang, Xiaolei and Tang, Xinyu and Zhao, Xin and Wang, Jingyuan and Wen, Ji-Rong", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.621", doi = "10.18653/v1/2023.emnlp-main.621", pages = "10052--10065", abstract = "The recent success of large language models (LLMs) has shown great potential to develop more powerful conversational recommender systems (CRSs), which rely on natural language conversations to satisfy user needs. In this paper, we embark on an investigation into the utilization of ChatGPT for CRSs, revealing the inadequacy of the existing evaluation protocol. It might overemphasize the matching with ground-truth items annotated by humans while neglecting the interactive nature of CRSs. To overcome the limitation, we further propose an **i**nteractive **Eva**luation approach based on **L**L**M**s, named **iEvaLM**, which harnesses LLM-based user simulators. Our evaluation approach can simulate various system-user interaction scenarios. Through the experiments on two public CRS datasets, we demonstrate notable improvements compared to the prevailing evaluation protocol. Furthermore, we emphasize the evaluation of explainability, and ChatGPT showcases persuasive explanation generation for its recommendations. Our study contributes to a deeper comprehension of the untapped potential of LLMs for CRSs and provides a more flexible and realistic evaluation approach for future research about LLM-based CRSs.", }

The recent success of large language models (LLMs) has shown great potential to develop more powerful conversational recommender systems (CRSs), which rely on natural language conversations to satisfy user needs. In this paper, we embark on an investigation into the utilization of ChatGPT for CRSs, revealing the inadequacy of the existing evaluation protocol. It might overemphasize the matching with ground-truth items annotated by humans while neglecting the interactive nature of CRSs. To overcome the limitation, we further propose an **i**nteractive **Eva**luation approach based on **L**L**M**s, named **iEvaLM**, which harnesses LLM-based user simulators. Our evaluation approach can simulate various system-user interaction scenarios. Through the experiments on two public CRS datasets, we demonstrate notable improvements compared to the prevailing evaluation protocol. Furthermore, we emphasize the evaluation of explainability, and ChatGPT showcases persuasive explanation generation for its recommendations. Our study contributes to a deeper comprehension of the untapped potential of LLMs for CRSs and provides a more flexible and realistic evaluation approach for future research about LLM-based CRSs.

Rethinking the Evaluation for Conversational Recommendation in the Era of Large Language Models

emnlp-main.621

2305.13112

Poster

https://aclanthology.org/2023.emnlp-main.622.bib

https://aclanthology.org/2023.emnlp-main.622/

@inproceedings{prasad-etal-2023-receval, title = "{R}e{CE}val: Evaluating Reasoning Chains via Correctness and Informativeness", author = "Prasad, Archiki and Saha, Swarnadeep and Zhou, Xiang and Bansal, Mohit", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.622", doi = "10.18653/v1/2023.emnlp-main.622", pages = "10066--10086", abstract = "Multi-step reasoning ability is fundamental to many natural language tasks, yet it is unclear what constitutes a good reasoning chain and how to evaluate them. Most existing methods focus solely on whether the reasoning chain leads to the correct conclusion, but this answer-oriented view may confound reasoning quality with other spurious shortcuts to predict the answer. To bridge this gap, we evaluate reasoning chains by viewing them as informal proofs that derive the final answer. Specifically, we propose ReCEval (Reasoning Chain Evaluation), a framework that evaluates reasoning chains via two key properties: (1) correctness, i.e., each step makes a valid inference based on information contained within the step, preceding steps, and input context, and (2) informativeness, i.e., each step provides new information that is helpful towards deriving the generated answer. We evaluate these properties by developing metrics using natural language inference models and $\mathcal{V}$-Information. On multiple datasets, we show that ReCEval effectively identifies various error types and yields notable improvements compared to prior methods. We analyze the impact of step boundaries, and previous steps on evaluating correctness and demonstrate that our informativeness metric captures the expected flow of information in high-quality reasoning chains. Finally, we show that scoring reasoning chains based on ReCEval improves downstream task performance.", }

Multi-step reasoning ability is fundamental to many natural language tasks, yet it is unclear what constitutes a good reasoning chain and how to evaluate them. Most existing methods focus solely on whether the reasoning chain leads to the correct conclusion, but this answer-oriented view may confound reasoning quality with other spurious shortcuts to predict the answer. To bridge this gap, we evaluate reasoning chains by viewing them as informal proofs that derive the final answer. Specifically, we propose ReCEval (Reasoning Chain Evaluation), a framework that evaluates reasoning chains via two key properties: (1) correctness, i.e., each step makes a valid inference based on information contained within the step, preceding steps, and input context, and (2) informativeness, i.e., each step provides new information that is helpful towards deriving the generated answer. We evaluate these properties by developing metrics using natural language inference models and $\mathcal{V}$-Information. On multiple datasets, we show that ReCEval effectively identifies various error types and yields notable improvements compared to prior methods. We analyze the impact of step boundaries, and previous steps on evaluating correctness and demonstrate that our informativeness metric captures the expected flow of information in high-quality reasoning chains. Finally, we show that scoring reasoning chains based on ReCEval improves downstream task performance.

ReCEval: Evaluating Reasoning Chains via Correctness and Informativeness

emnlp-main.622

2304.10703

Poster

https://aclanthology.org/2023.emnlp-main.623.bib

https://aclanthology.org/2023.emnlp-main.623/

@inproceedings{askari-etal-2023-expand, title = "Expand, Highlight, Generate: {RL}-driven Document Generation for Passage Reranking", author = "Askari, Arian and Aliannejadi, Mohammad and Meng, Chuan and Kanoulas, Evangelos and Verberne, Suzan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.623", doi = "10.18653/v1/2023.emnlp-main.623", pages = "10087--10099", abstract = "Generating synthetic training data based on large language models (LLMs) for ranking models has gained attention recently. Prior studies use LLMs to build pseudo query-document pairs by generating synthetic queries from documents in a corpus. In this paper, we propose a new perspective of data augmentation: generating synthetic documents from queries. To achieve this, we propose DocGen, that consists of a three-step pipeline that utilizes the few-shot capabilities of LLMs. DocGen pipeline performs synthetic document generation by (i) expanding, (ii) highlighting the original query, and then (iii) generating a synthetic document that is likely to be relevant to the query. To further improve the relevance between generated synthetic documents and their corresponding queries, we propose DocGen-RL, which regards the estimated relevance of the document as a reward and leverages reinforcement learning (RL) to optimize DocGen pipeline. Extensive experiments demonstrate that DocGen pipeline and DocGen-RL significantly outperform existing state-of-theart data augmentation methods, such as InPars, indicating that our new perspective of generating documents leverages the capacity of LLMs in generating synthetic data more effectively. We release the code, generated data, and model checkpoints to foster research in this area.", }

Generating synthetic training data based on large language models (LLMs) for ranking models has gained attention recently. Prior studies use LLMs to build pseudo query-document pairs by generating synthetic queries from documents in a corpus. In this paper, we propose a new perspective of data augmentation: generating synthetic documents from queries. To achieve this, we propose DocGen, that consists of a three-step pipeline that utilizes the few-shot capabilities of LLMs. DocGen pipeline performs synthetic document generation by (i) expanding, (ii) highlighting the original query, and then (iii) generating a synthetic document that is likely to be relevant to the query. To further improve the relevance between generated synthetic documents and their corresponding queries, we propose DocGen-RL, which regards the estimated relevance of the document as a reward and leverages reinforcement learning (RL) to optimize DocGen pipeline. Extensive experiments demonstrate that DocGen pipeline and DocGen-RL significantly outperform existing state-of-theart data augmentation methods, such as InPars, indicating that our new perspective of generating documents leverages the capacity of LLMs in generating synthetic data more effectively. We release the code, generated data, and model checkpoints to foster research in this area.

Expand, Highlight, Generate: RL-driven Document Generation for Passage Reranking

emnlp-main.623

Poster

https://aclanthology.org/2023.emnlp-main.624.bib

https://aclanthology.org/2023.emnlp-main.624/

@inproceedings{oshika-etal-2023-transformer, title = "Transformer-based Live Update Generation for Soccer Matches from Microblog Posts", author = "Oshika, Masashi and Yamada, Kosuke and Sasano, Ryohei and Takeda, Koichi", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.624", doi = "10.18653/v1/2023.emnlp-main.624", pages = "10100--10106", abstract = "It has been known to be difficult to generate adequate sports updates from a sequence of vast amounts of diverse live tweets, although the live sports viewing experience with tweets is gaining the popularity. In this paper, we focus on soccer matches and work on building a system to generate live updates for soccer matches from tweets so that users can instantly grasp a match{'}s progress and enjoy the excitement of the match from raw tweets. Our proposed system is based on a large pre-trained language model and incorporates a mechanism to control the number of updates and a mechanism to reduce the redundancy of duplicate and similar updates.", }

It has been known to be difficult to generate adequate sports updates from a sequence of vast amounts of diverse live tweets, although the live sports viewing experience with tweets is gaining the popularity. In this paper, we focus on soccer matches and work on building a system to generate live updates for soccer matches from tweets so that users can instantly grasp a match{'}s progress and enjoy the excitement of the match from raw tweets. Our proposed system is based on a large pre-trained language model and incorporates a mechanism to control the number of updates and a mechanism to reduce the redundancy of duplicate and similar updates.

Transformer-based Live Update Generation for Soccer Matches from Microblog Posts

emnlp-main.624

2310.16368

Poster

https://aclanthology.org/2023.emnlp-main.625.bib

https://aclanthology.org/2023.emnlp-main.625/

@inproceedings{bejan-etal-2023-make, title = "Make Every Example Count: On the Stability and Utility of Self-Influence for Learning from Noisy {NLP} Datasets", author = "Bejan, Irina and Sokolov, Artem and Filippova, Katja", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.625", doi = "10.18653/v1/2023.emnlp-main.625", pages = "10107--10121", abstract = "Increasingly larger datasets have become a standard ingredient to advancing the state-of-the-art in NLP. However, data quality might have already become the bottleneck to unlock further gains. Given the diversity and the sizes of modern datasets, standard data filtering is not straight-forward to apply, because of the multifacetedness of the harmful data and elusiveness of filtering rules that would generalize across multiple tasks. We study the fitness of task-agnostic self-influence scores of training examples for data cleaning, analyze their efficacy in capturing naturally occurring outliers, and investigate to what extent self-influence based data cleaning can improve downstream performance in machine translation, question answering and text classification, building up on recent approaches to self-influence calculation and automated curriculum learning.", }

Increasingly larger datasets have become a standard ingredient to advancing the state-of-the-art in NLP. However, data quality might have already become the bottleneck to unlock further gains. Given the diversity and the sizes of modern datasets, standard data filtering is not straight-forward to apply, because of the multifacetedness of the harmful data and elusiveness of filtering rules that would generalize across multiple tasks. We study the fitness of task-agnostic self-influence scores of training examples for data cleaning, analyze their efficacy in capturing naturally occurring outliers, and investigate to what extent self-influence based data cleaning can improve downstream performance in machine translation, question answering and text classification, building up on recent approaches to self-influence calculation and automated curriculum learning.

Make Every Example Count: On the Stability and Utility of Self-Influence for Learning from Noisy NLP Datasets

emnlp-main.625

2302.13959

Poster

https://aclanthology.org/2023.emnlp-main.626.bib

https://aclanthology.org/2023.emnlp-main.626/

@inproceedings{yun-etal-2023-appraising, title = "Appraising the Potential Uses and Harms of {LLM}s for Medical Systematic Reviews", author = "Yun, Hye and Marshall, Iain and Trikalinos, Thomas and Wallace, Byron", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.626", doi = "10.18653/v1/2023.emnlp-main.626", pages = "10122--10139", abstract = "Medical systematic reviews play a vital role in healthcare decision making and policy. However, their production is time-consuming, limiting the availability of high-quality and up-to-date evidence summaries. Recent advancements in LLMs offer the potential to automatically generate literature reviews on demand, addressing this issue. However, LLMs sometimes generate inaccurate (and potentially misleading) texts by hallucination or omission. In healthcare, this can make LLMs unusable at best and dangerous at worst. We conducted 16 interviews with international systematic review experts to characterize the perceived utility and risks of LLMs in the specific context of medical evidence reviews. Experts indicated that LLMs can assist in the writing process by drafting summaries, generating templates, distilling information, and crosschecking information. They also raised concerns regarding confidently composed but inaccurate LLM outputs and other potential downstream harms, including decreased accountability and proliferation of low-quality reviews. Informed by this qualitative analysis, we identify criteria for rigorous evaluation of biomedical LLMs aligned with domain expert views.", }

Medical systematic reviews play a vital role in healthcare decision making and policy. However, their production is time-consuming, limiting the availability of high-quality and up-to-date evidence summaries. Recent advancements in LLMs offer the potential to automatically generate literature reviews on demand, addressing this issue. However, LLMs sometimes generate inaccurate (and potentially misleading) texts by hallucination or omission. In healthcare, this can make LLMs unusable at best and dangerous at worst. We conducted 16 interviews with international systematic review experts to characterize the perceived utility and risks of LLMs in the specific context of medical evidence reviews. Experts indicated that LLMs can assist in the writing process by drafting summaries, generating templates, distilling information, and crosschecking information. They also raised concerns regarding confidently composed but inaccurate LLM outputs and other potential downstream harms, including decreased accountability and proliferation of low-quality reviews. Informed by this qualitative analysis, we identify criteria for rigorous evaluation of biomedical LLMs aligned with domain expert views.

Appraising the Potential Uses and Harms of LLMs for Medical Systematic Reviews

emnlp-main.626

2305.11828

Poster

https://aclanthology.org/2023.emnlp-main.627.bib

https://aclanthology.org/2023.emnlp-main.627/

@inproceedings{yu-etal-2023-promptst, title = "{P}rompt{ST}: Abstract Prompt Learning for End-to-End Speech Translation", author = "Yu, Tengfei and Ding, Liang and Liu, Xuebo and Chen, Kehai and Zhang, Meishan and Tao, Dacheng and Zhang, Min", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.627", doi = "10.18653/v1/2023.emnlp-main.627", pages = "10140--10154", abstract = "An end-to-end speech-to-text (S2T) translation model is usually initialized from a pre-trained speech recognition encoder and a pre-trained text-to-text (T2T) translation decoder. Although this straightforward setting has been shown empirically successful, there do not exist clear answers to the research questions: 1) how are speech and text modalities fused in S2T model and 2) how to better fuse the two modalities? In this paper, we take the first step toward understanding the fusion of speech and text features in S2T model. We first design and release a 10GB linguistic probing benchmark, namely Speech-Senteval, to investigate the acoustic and linguistic behaviors of S2T models. Preliminary analysis reveals that the uppermost encoder layers of the S2T model can not learn linguistic knowledge efficiently, which is crucial for accurate translation. Based on the finding, we further propose a simple plug-in prompt-learning strategy on the uppermost encoder layers to broaden the abstract representation power of the encoder of S2T models. We call such a prompt-enhanced S2T model PromptST. Experimental results on four widely-used S2T datasets show that PromptST can deliver significant improvements over a strong baseline by capturing richer linguistic knowledge. Benchmarks, code, and scripts are freely available at https://github.com/ytf-philp/PromptST.", }

An end-to-end speech-to-text (S2T) translation model is usually initialized from a pre-trained speech recognition encoder and a pre-trained text-to-text (T2T) translation decoder. Although this straightforward setting has been shown empirically successful, there do not exist clear answers to the research questions: 1) how are speech and text modalities fused in S2T model and 2) how to better fuse the two modalities? In this paper, we take the first step toward understanding the fusion of speech and text features in S2T model. We first design and release a 10GB linguistic probing benchmark, namely Speech-Senteval, to investigate the acoustic and linguistic behaviors of S2T models. Preliminary analysis reveals that the uppermost encoder layers of the S2T model can not learn linguistic knowledge efficiently, which is crucial for accurate translation. Based on the finding, we further propose a simple plug-in prompt-learning strategy on the uppermost encoder layers to broaden the abstract representation power of the encoder of S2T models. We call such a prompt-enhanced S2T model PromptST. Experimental results on four widely-used S2T datasets show that PromptST can deliver significant improvements over a strong baseline by capturing richer linguistic knowledge. Benchmarks, code, and scripts are freely available at https://github.com/ytf-philp/PromptST.

PromptST: Abstract Prompt Learning for End-to-End Speech Translation

emnlp-main.627

Poster

https://aclanthology.org/2023.emnlp-main.628.bib

https://aclanthology.org/2023.emnlp-main.628/

@inproceedings{lotz-etal-2023-text, title = "Text Rendering Strategies for Pixel Language Models", author = "Lotz, Jonas and Salesky, Elizabeth and Rust, Phillip and Elliott, Desmond", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.628", doi = "10.18653/v1/2023.emnlp-main.628", pages = "10155--10172", abstract = "Pixel-based language models process text rendered as images, which allows them to handle any script, making them a promising approach to open vocabulary language modelling. However, recent approaches use text renderers that produce a large set of almost-equivalent input patches, which may prove sub-optimal for downstream tasks, due to redundancy in the input representations. In this paper, we investigate four approaches to rendering text in the PIXEL model (Rust et al., 2023), and find that simple character bigram rendering brings improved performance on sentence-level tasks without compromising performance on token-level or multilingual tasks. This new rendering strategy also makes it possible to train a more compact model with only 22M parameters that performs on par with the original 86M parameter model. Our analyses show that character bigram rendering leads to a consistently better model but with an anisotropic patch embedding space, driven by a patch frequency bias, highlighting the connections between image patch- and tokenization-based language models.", }

Pixel-based language models process text rendered as images, which allows them to handle any script, making them a promising approach to open vocabulary language modelling. However, recent approaches use text renderers that produce a large set of almost-equivalent input patches, which may prove sub-optimal for downstream tasks, due to redundancy in the input representations. In this paper, we investigate four approaches to rendering text in the PIXEL model (Rust et al., 2023), and find that simple character bigram rendering brings improved performance on sentence-level tasks without compromising performance on token-level or multilingual tasks. This new rendering strategy also makes it possible to train a more compact model with only 22M parameters that performs on par with the original 86M parameter model. Our analyses show that character bigram rendering leads to a consistently better model but with an anisotropic patch embedding space, driven by a patch frequency bias, highlighting the connections between image patch- and tokenization-based language models.

Text Rendering Strategies for Pixel Language Models

emnlp-main.628

2311.00522

https://huggingface.co/papers/2311.00522

Poster

https://aclanthology.org/2023.emnlp-main.629.bib

https://aclanthology.org/2023.emnlp-main.629/

@inproceedings{chowdhury-etal-2023-apollo, title = "{AP}o{LL}o : Unified Adapter and Prompt Learning for Vision Language Models", author = "Chowdhury, Sanjoy and Nag, Sayan and Manocha, Dinesh", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.629", doi = "10.18653/v1/2023.emnlp-main.629", pages = "10173--10187", abstract = "The choice of input text prompt plays a critical role in the performance of Vision-Language Pretrained (VLP) models such as CLIP. We present APoLLo, a unified multi-modal approach that combines Adapter and Prompt learning for Vision-Language models. Our method is designed to substantially improve the generalization capabilities of VLP models when they are fine-tuned in a few-shot setting. We introduce trainable cross-attention-based adapter layers in conjunction with vision and language encoders to strengthen the alignment between the two modalities. We enforce consistency between the respective encoder branches (receiving augmented inputs) to prevent overfitting in downstream tasks. Our method is evaluated on three representative tasks: generalization to novel classes, cross-dataset evaluation, and unseen domain shifts. In practice, APoLLo achieves a relative gain up to 6.03{\%} over MaPLe (SOTA) on novel classes for 10 diverse image recognition datasets.", }

The choice of input text prompt plays a critical role in the performance of Vision-Language Pretrained (VLP) models such as CLIP. We present APoLLo, a unified multi-modal approach that combines Adapter and Prompt learning for Vision-Language models. Our method is designed to substantially improve the generalization capabilities of VLP models when they are fine-tuned in a few-shot setting. We introduce trainable cross-attention-based adapter layers in conjunction with vision and language encoders to strengthen the alignment between the two modalities. We enforce consistency between the respective encoder branches (receiving augmented inputs) to prevent overfitting in downstream tasks. Our method is evaluated on three representative tasks: generalization to novel classes, cross-dataset evaluation, and unseen domain shifts. In practice, APoLLo achieves a relative gain up to 6.03{\%} over MaPLe (SOTA) on novel classes for 10 diverse image recognition datasets.

APoLLo : Unified Adapter and Prompt Learning for Vision Language Models

emnlp-main.629

Poster

https://aclanthology.org/2023.emnlp-main.630.bib

https://aclanthology.org/2023.emnlp-main.630/

@inproceedings{kang-shin-2023-samrank, title = "{SAMR}ank: Unsupervised Keyphrase Extraction using Self-Attention Map in {BERT} and {GPT}-2", author = "Kang, Byungha and Shin, Youhyun", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.630", doi = "10.18653/v1/2023.emnlp-main.630", pages = "10188--10201", abstract = "We propose a novel unsupervised keyphrase extraction approach, called SAMRank, which uses only a self-attention map in a pre-trained language model (PLM) to determine the importance of phrases. Most recent approaches for unsupervised keyphrase extraction mainly utilize contextualized embeddings to capture semantic relevance between words, sentences, and documents. However, due to the anisotropic nature of contextual embeddings, these approaches may not be optimal for semantic similarity measurements. SAMRank as proposed here computes the importance of phrases solely leveraging a self-attention map in a PLM, in this case BERT and GPT-2, eliminating the need to measure embedding similarities. To assess the level of importance, SAMRank combines both global and proportional attention scores through calculations using a self-attention map. We evaluate the SAMRank on three keyphrase extraction datasets: Inspec, SemEval2010, and SemEval2017. The experimental results show that SAMRank outperforms most embedding-based models on both long and short documents and demonstrating that it is possible to use only a self-attention map for keyphrase extraction without relying on embeddings. Source code is available at https://github.com/kangnlp/SAMRank.", }

We propose a novel unsupervised keyphrase extraction approach, called SAMRank, which uses only a self-attention map in a pre-trained language model (PLM) to determine the importance of phrases. Most recent approaches for unsupervised keyphrase extraction mainly utilize contextualized embeddings to capture semantic relevance between words, sentences, and documents. However, due to the anisotropic nature of contextual embeddings, these approaches may not be optimal for semantic similarity measurements. SAMRank as proposed here computes the importance of phrases solely leveraging a self-attention map in a PLM, in this case BERT and GPT-2, eliminating the need to measure embedding similarities. To assess the level of importance, SAMRank combines both global and proportional attention scores through calculations using a self-attention map. We evaluate the SAMRank on three keyphrase extraction datasets: Inspec, SemEval2010, and SemEval2017. The experimental results show that SAMRank outperforms most embedding-based models on both long and short documents and demonstrating that it is possible to use only a self-attention map for keyphrase extraction without relying on embeddings. Source code is available at https://github.com/kangnlp/SAMRank.

SAMRank: Unsupervised Keyphrase Extraction using Self-Attention Map in BERT and GPT-2

emnlp-main.630

Poster

https://aclanthology.org/2023.emnlp-main.631.bib

https://aclanthology.org/2023.emnlp-main.631/

@inproceedings{ishii-etal-2023-contrastive, title = "Contrastive Learning for Inference in Dialogue", author = "Ishii, Etsuko and Xu, Yan and Wilie, Bryan and Ji, Ziwei and Lovenia, Holy and Chung, Willy and Fung, Pascale", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.631", doi = "10.18653/v1/2023.emnlp-main.631", pages = "10202--10221", abstract = "Inference, especially those derived from inductive processes, is a crucial component in our conversation to complement the information implicitly or explicitly conveyed by a speaker. While recent large language models show remarkable advances in inference tasks, their performance in inductive reasoning, where not all information is present in the context, is far behind deductive reasoning. In this paper, we analyze the behavior of the models based on the task difficulty defined by the semantic information gap {--} which distinguishes inductive and deductive reasoning. Our analysis reveals that the information gap between dialogue contexts and desired inferences renders the inductive inference process more challenging. To mitigate this information gap, we investigate a contrastive learning approach by feeding negative samples. Our experiments suggest negative samples help models understand what is wrong and improve their inference generations.", }

Inference, especially those derived from inductive processes, is a crucial component in our conversation to complement the information implicitly or explicitly conveyed by a speaker. While recent large language models show remarkable advances in inference tasks, their performance in inductive reasoning, where not all information is present in the context, is far behind deductive reasoning. In this paper, we analyze the behavior of the models based on the task difficulty defined by the semantic information gap {--} which distinguishes inductive and deductive reasoning. Our analysis reveals that the information gap between dialogue contexts and desired inferences renders the inductive inference process more challenging. To mitigate this information gap, we investigate a contrastive learning approach by feeding negative samples. Our experiments suggest negative samples help models understand what is wrong and improve their inference generations.

Contrastive Learning for Inference in Dialogue

emnlp-main.631

Poster

https://aclanthology.org/2023.emnlp-main.632.bib

https://aclanthology.org/2023.emnlp-main.632/

@inproceedings{yao-etal-2023-editing, title = "Editing Large Language Models: Problems, Methods, and Opportunities", author = "Yao, Yunzhi and Wang, Peng and Tian, Bozhong and Cheng, Siyuan and Li, Zhoubo and Deng, Shumin and Chen, Huajun and Zhang, Ningyu", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.632", doi = "10.18653/v1/2023.emnlp-main.632", pages = "10222--10240", abstract = "Despite the ability to train capable LLMs, the methodology for maintaining their relevancy and rectifying errors remains elusive. To this end, the past few years have witnessed a surge in techniques for editing LLMs, the objective of which is to alter the behavior of LLMs \textbf{efficiently} within a specific domain without negatively impacting performance across other inputs. This paper embarks on a deep exploration of the problems, methods, and opportunities related to model editing for LLMs. In particular, we provide an exhaustive overview of the task definition and challenges associated with model editing, along with an in-depth empirical analysis of the most progressive methods currently at our disposal. We also build a new benchmark dataset to facilitate a more robust evaluation and pinpoint enduring issues intrinsic to existing techniques. Our objective is to provide valuable insights into the effectiveness and feasibility of each editing technique, thereby assisting the community in making informed decisions on the selection of the most appropriate method for a specific task or context.", }

Despite the ability to train capable LLMs, the methodology for maintaining their relevancy and rectifying errors remains elusive. To this end, the past few years have witnessed a surge in techniques for editing LLMs, the objective of which is to alter the behavior of LLMs \textbf{efficiently} within a specific domain without negatively impacting performance across other inputs. This paper embarks on a deep exploration of the problems, methods, and opportunities related to model editing for LLMs. In particular, we provide an exhaustive overview of the task definition and challenges associated with model editing, along with an in-depth empirical analysis of the most progressive methods currently at our disposal. We also build a new benchmark dataset to facilitate a more robust evaluation and pinpoint enduring issues intrinsic to existing techniques. Our objective is to provide valuable insights into the effectiveness and feasibility of each editing technique, thereby assisting the community in making informed decisions on the selection of the most appropriate method for a specific task or context.

Editing Large Language Models: Problems, Methods, and Opportunities

emnlp-main.632

Poster

https://aclanthology.org/2023.emnlp-main.633.bib

https://aclanthology.org/2023.emnlp-main.633/

@inproceedings{huang-etal-2023-markqa, title = "{M}ark{QA}: A large scale {KBQA} dataset with numerical reasoning", author = "Huang, Xiang and Cheng, Sitao and Bao, Yuheng and Huang, Shanshan and Qu, Yuzhong", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.633", pages = "10241--10259", abstract = "While question answering over knowledge bases (KBQA) has shown progress in addressing factoid questions, KBQA with numerical reasoning remains relatively unexplored. In this paper, we focus on the complex numerical reasoning in KBQA, and propose a new task, NR-KBQA, which necessitates the ability to perform both multi-hop reasoning and numerical reasoning. We also design a logic form in Python format called PyQL to represent the reasoning process of numerical reasoning questions. To facilitate the development of NR-KBQA, we present a large NR-KBQA dataset called MarkQA, which is automatically constructed by a small set of seeds. Each question in MarkQA is annotated with its corresponding SPARQL query, alongside the step-by-step reasoning path in the QDMR format and PyQL program. Experimental results of some state-of-the-art QA methods performed on the MarkQA dataset show that complex numerical reasoning in KBQA faces great challenges.", }

While question answering over knowledge bases (KBQA) has shown progress in addressing factoid questions, KBQA with numerical reasoning remains relatively unexplored. In this paper, we focus on the complex numerical reasoning in KBQA, and propose a new task, NR-KBQA, which necessitates the ability to perform both multi-hop reasoning and numerical reasoning. We also design a logic form in Python format called PyQL to represent the reasoning process of numerical reasoning questions. To facilitate the development of NR-KBQA, we present a large NR-KBQA dataset called MarkQA, which is automatically constructed by a small set of seeds. Each question in MarkQA is annotated with its corresponding SPARQL query, alongside the step-by-step reasoning path in the QDMR format and PyQL program. Experimental results of some state-of-the-art QA methods performed on the MarkQA dataset show that complex numerical reasoning in KBQA faces great challenges.

MarkQA: A large scale KBQA dataset with numerical reasoning

emnlp-main.633

2310.15517

https://huggingface.co/papers/2310.15517

Poster

https://aclanthology.org/2023.emnlp-main.634.bib

https://aclanthology.org/2023.emnlp-main.634/

@inproceedings{levy-etal-2023-comparing, title = "Comparing Biases and the Impact of Multilingual Training across Multiple Languages", author = "Levy, Sharon and John, Neha and Liu, Ling and Vyas, Yogarshi and Ma, Jie and Fujinuma, Yoshinari and Ballesteros, Miguel and Castelli, Vittorio and Roth, Dan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.634", doi = "10.18653/v1/2023.emnlp-main.634", pages = "10260--10280", abstract = "Studies in bias and fairness in natural language processing have primarily examined social biases within a single language and/or across few attributes (e.g. gender, race). However, biases can manifest differently across various languages for individual attributes. As a result, it is critical to examine biases within each language and attribute. Of equal importance is to study how these biases compare across languages and how the biases are affected when training a model on multilingual data versus monolingual data. We present a bias analysis across Italian, Chinese, English, Hebrew, and Spanish on the downstream sentiment analysis task to observe whether specific demographics are viewed more positively. We study bias similarities and differences across these languages and investigate the impact of multilingual vs. monolingual training data. We adapt existing sentiment bias templates in English to Italian, Chinese, Hebrew, and Spanish for four attributes: race, religion, nationality, and gender. Our results reveal similarities in bias expression such as favoritism of groups that are dominant in each language{'}s culture (e.g. majority religions and nationalities). Additionally, we find an increased variation in predictions across protected groups, indicating bias amplification, after multilingual finetuning in comparison to multilingual pretraining.", }

Studies in bias and fairness in natural language processing have primarily examined social biases within a single language and/or across few attributes (e.g. gender, race). However, biases can manifest differently across various languages for individual attributes. As a result, it is critical to examine biases within each language and attribute. Of equal importance is to study how these biases compare across languages and how the biases are affected when training a model on multilingual data versus monolingual data. We present a bias analysis across Italian, Chinese, English, Hebrew, and Spanish on the downstream sentiment analysis task to observe whether specific demographics are viewed more positively. We study bias similarities and differences across these languages and investigate the impact of multilingual vs. monolingual training data. We adapt existing sentiment bias templates in English to Italian, Chinese, Hebrew, and Spanish for four attributes: race, religion, nationality, and gender. Our results reveal similarities in bias expression such as favoritism of groups that are dominant in each language{'}s culture (e.g. majority religions and nationalities). Additionally, we find an increased variation in predictions across protected groups, indicating bias amplification, after multilingual finetuning in comparison to multilingual pretraining.

Comparing Biases and the Impact of Multilingual Training across Multiple Languages

emnlp-main.634

2305.11242

Poster

https://aclanthology.org/2023.emnlp-main.635.bib

https://aclanthology.org/2023.emnlp-main.635/

@inproceedings{qian-etal-2023-hutcrs, title = "{H}ut{CRS}: Hierarchical User-Interest Tracking for Conversational Recommender System", author = "Qian, Mingjie and Zheng, Yongsen and Qin, Jinghui and Lin, Liang", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.635", doi = "10.18653/v1/2023.emnlp-main.635", pages = "10281--10290", abstract = "Conversational Recommender System (CRS) aims to explicitly acquire user preferences towards items and attributes through natural language conversations. However, existing CRS methods ask users to provide explicit answers (yes/no) for each attribute they require, regardless of users{'} knowledge or interest, which may significantly reduce the user experience and semantic consistency. Furthermore, these methods assume that users like all attributes of the target item and dislike those unrelated to it, which can introduce bias in attribute-level feedback and impede the system{'}s ability to accurately identify the target item. To address these issues, we propose a more realistic, user-friendly, and explainable CRS framework called Hierarchical User-Interest Tracking for Conversational Recommender System (HutCRS). HutCRS portrays the conversation as a hierarchical interest tree that consists of two stages. In stage I, the system identifies the aspects that the user prefers while the system asks about attributes related to these positive aspects or recommends items in stage II. In addition, we develop a Hierarchical-Interest Policy Learning (HIPL) module to integrate the decision-making process of which aspects to ask and when to ask about attributes or recommend items. Moreover, we classify the attribute-level feedback results to further enhance the system{'}s ability to capture special information, such as attribute instances that are accepted by users but not presented in their historical interactive data. Extensive experiments on four benchmark datasets demonstrate the superiority of our method. The implementation of HutCRS is publicly available at https://github.com/xinle1129/HutCRS.", }

Conversational Recommender System (CRS) aims to explicitly acquire user preferences towards items and attributes through natural language conversations. However, existing CRS methods ask users to provide explicit answers (yes/no) for each attribute they require, regardless of users{'} knowledge or interest, which may significantly reduce the user experience and semantic consistency. Furthermore, these methods assume that users like all attributes of the target item and dislike those unrelated to it, which can introduce bias in attribute-level feedback and impede the system{'}s ability to accurately identify the target item. To address these issues, we propose a more realistic, user-friendly, and explainable CRS framework called Hierarchical User-Interest Tracking for Conversational Recommender System (HutCRS). HutCRS portrays the conversation as a hierarchical interest tree that consists of two stages. In stage I, the system identifies the aspects that the user prefers while the system asks about attributes related to these positive aspects or recommends items in stage II. In addition, we develop a Hierarchical-Interest Policy Learning (HIPL) module to integrate the decision-making process of which aspects to ask and when to ask about attributes or recommend items. Moreover, we classify the attribute-level feedback results to further enhance the system{'}s ability to capture special information, such as attribute instances that are accepted by users but not presented in their historical interactive data. Extensive experiments on four benchmark datasets demonstrate the superiority of our method. The implementation of HutCRS is publicly available at https://github.com/xinle1129/HutCRS.

HutCRS: Hierarchical User-Interest Tracking for Conversational Recommender System

emnlp-main.635

Poster

https://aclanthology.org/2023.emnlp-main.636.bib

https://aclanthology.org/2023.emnlp-main.636/

@inproceedings{song-etal-2023-large, title = "Large Language Models Meet Open-World Intent Discovery and Recognition: An Evaluation of {C}hat{GPT}", author = "Song, Xiaoshuai and He, Keqing and Wang, Pei and Dong, Guanting and Mou, Yutao and Wang, Jingang and Xian, Yunsen and Cai, Xunliang and Xu, Weiran", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.636", doi = "10.18653/v1/2023.emnlp-main.636", pages = "10291--10304", abstract = "The tasks of out-of-domain (OOD) intent discovery and generalized intent discovery (GID) aim to extend a closed intent classifier to open-world intent sets, which is crucial to task-oriented dialogue (TOD) systems. Previous methods address them by fine-tuning discriminative models. Recently, although some studies has been exploring the application of large language models (LLMs) represented by ChatGPT to various downstream tasks, it is still unclear for the ability of ChatGPT to discover and incrementally extent OOD intents. In this paper, we comprehensively evaluate ChatGPT on OOD intent discovery and GID, and then outline the strengths and weaknesses of ChatGPT. Overall, ChatGPT exhibits consistent advantages under zero-shot settings, but is still at a disadvantage compared to fine-tuned models. More deeply, through a series of analytical experiments, we summarize and discuss the challenges faced by LLMs including clustering, domain-specific understanding, and cross-domain in-context learning scenarios. Finally, we provide empirical guidance for future directions to address these challenges.", }

The tasks of out-of-domain (OOD) intent discovery and generalized intent discovery (GID) aim to extend a closed intent classifier to open-world intent sets, which is crucial to task-oriented dialogue (TOD) systems. Previous methods address them by fine-tuning discriminative models. Recently, although some studies has been exploring the application of large language models (LLMs) represented by ChatGPT to various downstream tasks, it is still unclear for the ability of ChatGPT to discover and incrementally extent OOD intents. In this paper, we comprehensively evaluate ChatGPT on OOD intent discovery and GID, and then outline the strengths and weaknesses of ChatGPT. Overall, ChatGPT exhibits consistent advantages under zero-shot settings, but is still at a disadvantage compared to fine-tuned models. More deeply, through a series of analytical experiments, we summarize and discuss the challenges faced by LLMs including clustering, domain-specific understanding, and cross-domain in-context learning scenarios. Finally, we provide empirical guidance for future directions to address these challenges.

Large Language Models Meet Open-World Intent Discovery and Recognition: An Evaluation of ChatGPT

emnlp-main.636

2310.10176

https://huggingface.co/papers/2310.10176

Poster

https://aclanthology.org/2023.emnlp-main.637.bib

https://aclanthology.org/2023.emnlp-main.637/

@inproceedings{chiang-yogatama-2023-distributional, title = "The Distributional Hypothesis Does Not Fully Explain the Benefits of Masked Language Model Pretraining", author = "Chiang, Ting-Rui and Yogatama, Dani", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.637", doi = "10.18653/v1/2023.emnlp-main.637", pages = "10305--10321", abstract = "We analyze the masked language modeling pretraining objective function from the perspective of the Distributional Hypothesis. We investigate whether the better sample efficiency and the better generalization capability of models pretrained with masked language modeling can be attributed to the semantic similarity encoded in the pretraining data{'}s distributional property. Via a synthetic dataset, our analysis suggests that distributional property indeed leads to the better sample efficiency of pretrained masked language models, but does not fully explain the generalization capability. We also conduct an analysis over two real-world datasets and demonstrate that the distributional property does not explain the generalization ability of pretrained natural language models either. Our results illustrate our limited understanding of model pretraining and provide future research directions.", }

We analyze the masked language modeling pretraining objective function from the perspective of the Distributional Hypothesis. We investigate whether the better sample efficiency and the better generalization capability of models pretrained with masked language modeling can be attributed to the semantic similarity encoded in the pretraining data{'}s distributional property. Via a synthetic dataset, our analysis suggests that distributional property indeed leads to the better sample efficiency of pretrained masked language models, but does not fully explain the generalization capability. We also conduct an analysis over two real-world datasets and demonstrate that the distributional property does not explain the generalization ability of pretrained natural language models either. Our results illustrate our limited understanding of model pretraining and provide future research directions.

The Distributional Hypothesis Does Not Fully Explain the Benefits of Masked Language Model Pretraining

emnlp-main.637

2310.16261

Poster

https://aclanthology.org/2023.emnlp-main.638.bib

https://aclanthology.org/2023.emnlp-main.638/

@inproceedings{yu-etal-2023-simple, title = "Simple and Effective Input Reformulations for Translation", author = "Yu, Brian and Lillemark, Hansen and Keutzer, Kurt", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.638", doi = "10.18653/v1/2023.emnlp-main.638", pages = "10322--10334", abstract = "Foundation language models learn from their finetuning input context in different ways. In this paper, we reformulate inputs during finetuning for challenging translation tasks, leveraging model strengths from pretraining in novel ways to improve downstream performance. These reformulations are simple data level modifications, require no additional collection of training data or modification of data at inference time. They can be applied either on single language pair translation tasks or massively multilingual translation tasks. Experiments with these techniques demonstrate significant performance improvements up to \textbf{3.5 chrF++ on the Flores200 translation benchmark}. We hope our research accessibly improves finetuning data efficiency, enabling more effective training to scalably improve state-of-the-art performance. Our code is released here.", }

Foundation language models learn from their finetuning input context in different ways. In this paper, we reformulate inputs during finetuning for challenging translation tasks, leveraging model strengths from pretraining in novel ways to improve downstream performance. These reformulations are simple data level modifications, require no additional collection of training data or modification of data at inference time. They can be applied either on single language pair translation tasks or massively multilingual translation tasks. Experiments with these techniques demonstrate significant performance improvements up to \textbf{3.5 chrF++ on the Flores200 translation benchmark}. We hope our research accessibly improves finetuning data efficiency, enabling more effective training to scalably improve state-of-the-art performance. Our code is released here.

Simple and Effective Input Reformulations for Translation

emnlp-main.638

2311.06696

Poster

https://aclanthology.org/2023.emnlp-main.639.bib

https://aclanthology.org/2023.emnlp-main.639/

@inproceedings{nandwani-etal-2023-pointwise, title = "Pointwise Mutual Information Based Metric and Decoding Strategy for Faithful Generation in Document Grounded Dialogs", author = "Nandwani, Yatin and Kumar, Vineet and Raghu, Dinesh and Joshi, Sachindra and Lastras, Luis", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.639", doi = "10.18653/v1/2023.emnlp-main.639", pages = "10335--10347", abstract = "A major concern in using deep learning based generative models for document-grounded dialogs is the potential generation of responses that are not faithful to the underlying document. Existing automated metrics used for evaluating the faithfulness of response with respect to the grounding document measure the degree of similarity between the generated response and the document{'}s content. However, these automated metrics are far from being well aligned with human judgments. Therefore, to improve the measurement of faithfulness, we propose a new metric that utilizes (Conditional) Point-wise Mutual Information (PMI) between the generated response and the source document, conditioned on the dialogue. PMI quantifies the extent to which the document influences the generated response {--} with a higher PMI indicating a more faithful response. We build upon this idea to create a new decoding technique that incorporates PMI into the response generation process to predict more faithful responses. Our experiments on the BEGIN benchmark demonstrate an improved correlation of our metric with human evaluation. We also show that our decoding technique is effective in generating more faithful responses when compared to standard decoding techniques on a set of publicly available document-grounded dialog datasets.", }

A major concern in using deep learning based generative models for document-grounded dialogs is the potential generation of responses that are not faithful to the underlying document. Existing automated metrics used for evaluating the faithfulness of response with respect to the grounding document measure the degree of similarity between the generated response and the document{'}s content. However, these automated metrics are far from being well aligned with human judgments. Therefore, to improve the measurement of faithfulness, we propose a new metric that utilizes (Conditional) Point-wise Mutual Information (PMI) between the generated response and the source document, conditioned on the dialogue. PMI quantifies the extent to which the document influences the generated response {--} with a higher PMI indicating a more faithful response. We build upon this idea to create a new decoding technique that incorporates PMI into the response generation process to predict more faithful responses. Our experiments on the BEGIN benchmark demonstrate an improved correlation of our metric with human evaluation. We also show that our decoding technique is effective in generating more faithful responses when compared to standard decoding techniques on a set of publicly available document-grounded dialog datasets.

Pointwise Mutual Information Based Metric and Decoding Strategy for Faithful Generation in Document Grounded Dialogs

emnlp-main.639

2305.12191

Poster

https://aclanthology.org/2023.emnlp-main.640.bib

https://aclanthology.org/2023.emnlp-main.640/

@inproceedings{rohatgi-etal-2023-acl, title = "The {ACL} {OCL} Corpus: Advancing Open Science in Computational Linguistics", author = "Rohatgi, Shaurya and Qin, Yanxia and Aw, Benjamin and Unnithan, Niranjana and Kan, Min-Yen", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.640", doi = "10.18653/v1/2023.emnlp-main.640", pages = "10348--10361", abstract = "We present ACL OCL, a scholarly corpus derived from the ACL Anthology to assist Open scientific research in the Computational Linguistics domain. Integrating and enhancing the previous versions of the ACL Anthology, the ACL OCL contributes metadata, PDF files, citation graphs and additional structured full texts with sections, figures, and links to a large knowledge resource (Semantic Scholar). The ACL OCL spans seven decades, containing 73K papers, alongside 210K figures. We spotlight how ACL OCL applies to observe trends in computational linguistics. By detecting paper topics with a supervised neural model, we note that interest in {``}Syntax: Tagging, Chunking and Parsing{''} is waning and {``}Natural Language Generation{''} is resurging. Our dataset is available from HuggingFace (https://huggingface.co/datasets/WINGNUS/ACL-OCL).", }

We present ACL OCL, a scholarly corpus derived from the ACL Anthology to assist Open scientific research in the Computational Linguistics domain. Integrating and enhancing the previous versions of the ACL Anthology, the ACL OCL contributes metadata, PDF files, citation graphs and additional structured full texts with sections, figures, and links to a large knowledge resource (Semantic Scholar). The ACL OCL spans seven decades, containing 73K papers, alongside 210K figures. We spotlight how ACL OCL applies to observe trends in computational linguistics. By detecting paper topics with a supervised neural model, we note that interest in {``}Syntax: Tagging, Chunking and Parsing{''} is waning and {``}Natural Language Generation{''} is resurging. Our dataset is available from HuggingFace (https://huggingface.co/datasets/WINGNUS/ACL-OCL).

The ACL OCL Corpus: Advancing Open Science in Computational Linguistics

emnlp-main.640

2305.14996

https://huggingface.co/papers/2305.14996

Poster

https://aclanthology.org/2023.emnlp-main.641.bib

https://aclanthology.org/2023.emnlp-main.641/

@inproceedings{conti-wisniewski-2023-using, title = "Using Artificial {F}rench Data to Understand the Emergence of Gender Bias in Transformer Language Models", author = "Conti, Lina and Wisniewski, Guillaume", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.641", doi = "10.18653/v1/2023.emnlp-main.641", pages = "10362--10371", abstract = "Numerous studies have demonstrated the ability of neural language models to learn various linguistic properties without direct supervision. This work takes an initial step towards exploring the less researched topic of how neural models discover linguistic properties of words, such as gender, as well as the rules governing their usage. We propose to use an artificial corpus generated by a PCFG based on French to precisely control the gender distribution in the training data and determine under which conditions a model correctly captures gender information or, on the contrary, appears gender-biased.", }

Numerous studies have demonstrated the ability of neural language models to learn various linguistic properties without direct supervision. This work takes an initial step towards exploring the less researched topic of how neural models discover linguistic properties of words, such as gender, as well as the rules governing their usage. We propose to use an artificial corpus generated by a PCFG based on French to precisely control the gender distribution in the training data and determine under which conditions a model correctly captures gender information or, on the contrary, appears gender-biased.

Using Artificial French Data to Understand the Emergence of Gender Bias in Transformer Language Models

emnlp-main.641

2310.15852

Poster

https://aclanthology.org/2023.emnlp-main.642.bib

https://aclanthology.org/2023.emnlp-main.642/

@inproceedings{amalvy-etal-2023-learning, title = "Learning to Rank Context for Named Entity Recognition Using a Synthetic Dataset", author = "Amalvy, Arthur and Labatut, Vincent and Dufour, Richard", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.642", doi = "10.18653/v1/2023.emnlp-main.642", pages = "10372--10382", abstract = "While recent pre-trained transformer-based models can perform named entity recognition (NER) with great accuracy, their limited range remains an issue when applied to long documents such as whole novels. To alleviate this issue, a solution is to retrieve relevant context at the document level. Unfortunately, the lack of supervision for such a task means one has to settle for unsupervised approaches. Instead, we propose to generate a synthetic context retrieval training dataset using Alpaca, an instruction-tuned large language model (LLM). Using this dataset, we train a neural context retriever based on a BERT model that is able to find relevant context for NER. We show that our method outperforms several retrieval baselines for the NER task on an English literary dataset composed of the first chapter of 40 books.", }

While recent pre-trained transformer-based models can perform named entity recognition (NER) with great accuracy, their limited range remains an issue when applied to long documents such as whole novels. To alleviate this issue, a solution is to retrieve relevant context at the document level. Unfortunately, the lack of supervision for such a task means one has to settle for unsupervised approaches. Instead, we propose to generate a synthetic context retrieval training dataset using Alpaca, an instruction-tuned large language model (LLM). Using this dataset, we train a neural context retriever based on a BERT model that is able to find relevant context for NER. We show that our method outperforms several retrieval baselines for the NER task on an English literary dataset composed of the first chapter of 40 books.

Learning to Rank Context for Named Entity Recognition Using a Synthetic Dataset

emnlp-main.642

2310.10118

Poster

https://aclanthology.org/2023.emnlp-main.643.bib

https://aclanthology.org/2023.emnlp-main.643/

@inproceedings{lahoti-etal-2023-improving, title = "Improving Diversity of Demographic Representation in Large Language Models via Collective-Critiques and Self-Voting", author = "Lahoti, Preethi and Blumm, Nicholas and Ma, Xiao and Kotikalapudi, Raghavendra and Potluri, Sahitya and Tan, Qijun and Srinivasan, Hansa and Packer, Ben and Beirami, Ahmad and Beutel, Alex and Chen, Jilin", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.643", doi = "10.18653/v1/2023.emnlp-main.643", pages = "10383--10405", abstract = "A crucial challenge for generative large language models (LLMs) is diversity: when a user{'}s prompt is under-specified, models may follow implicit assumptions while generating a response, which may result in homogenization of the responses, as well as certain demographic groups being under-represented or even erased from the generated responses. In this paper, we formalize the problem diversity of representation in LLM generations. We present evaluation datasets and propose metrics to measure diversity in generated responses along people and culture axes. We find that LLMs understand the notion of diversity, and that they can reason and critique their own responses for that goal. This finding motivated a new prompting technique called collective-critique and self-voting (CCSV) to self-improve people diversity of LLMs by tapping into its diversity reasoning capabilities, without relying on handcrafted examples or prompt tuning. Extensive empirical experiments with both human and automated evaluations show that our proposed approach is effective at improving people and culture diversity, and outperforms all baseline methods by a large margin.", }

A crucial challenge for generative large language models (LLMs) is diversity: when a user{'}s prompt is under-specified, models may follow implicit assumptions while generating a response, which may result in homogenization of the responses, as well as certain demographic groups being under-represented or even erased from the generated responses. In this paper, we formalize the problem diversity of representation in LLM generations. We present evaluation datasets and propose metrics to measure diversity in generated responses along people and culture axes. We find that LLMs understand the notion of diversity, and that they can reason and critique their own responses for that goal. This finding motivated a new prompting technique called collective-critique and self-voting (CCSV) to self-improve people diversity of LLMs by tapping into its diversity reasoning capabilities, without relying on handcrafted examples or prompt tuning. Extensive empirical experiments with both human and automated evaluations show that our proposed approach is effective at improving people and culture diversity, and outperforms all baseline methods by a large margin.

Improving Diversity of Demographic Representation in Large Language Models via Collective-Critiques and Self-Voting

emnlp-main.643

Poster

https://aclanthology.org/2023.emnlp-main.644.bib

https://aclanthology.org/2023.emnlp-main.644/

@inproceedings{peng-etal-2023-hidding, title = "Hidding the Ghostwriters: An Adversarial Evaluation of {AI}-Generated Student Essay Detection", author = "Peng, Xinlin and Zhou, Ying and He, Ben and Sun, Le and Sun, Yingfei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.644", doi = "10.18653/v1/2023.emnlp-main.644", pages = "10406--10419", abstract = "Large language models (LLMs) have exhibited remarkable capabilities in text generation tasks. However, the utilization of these models carries inherent risks, including but not limited to plagiarism, the dissemination of fake news, and issues in educational exercises. Although several detectors have been proposed to address these concerns, their effectiveness against adversarial perturbations, specifically in the context of student essay writing, remains largely unexplored. This paper aims to bridge this gap by constructing AIG-ASAP, an AI-generated student essay dataset, employing a range of text perturbation methods that are expected to generate high-quality essays while evading detection. Through empirical experiments, we assess the performance of current AIGC detectors on the AIG-ASAP dataset. The results reveal that the existing detectors can be easily circumvented using straightforward automatic adversarial attacks. Specifically, we explore word substitution and sentence substitution perturbation methods that effectively evade detection while maintaining the quality of the generated essays. This highlights the urgent need for more accurate and robust methods to detect AI-generated student essays in the education domain. Code and data are released for public use.", }

Large language models (LLMs) have exhibited remarkable capabilities in text generation tasks. However, the utilization of these models carries inherent risks, including but not limited to plagiarism, the dissemination of fake news, and issues in educational exercises. Although several detectors have been proposed to address these concerns, their effectiveness against adversarial perturbations, specifically in the context of student essay writing, remains largely unexplored. This paper aims to bridge this gap by constructing AIG-ASAP, an AI-generated student essay dataset, employing a range of text perturbation methods that are expected to generate high-quality essays while evading detection. Through empirical experiments, we assess the performance of current AIGC detectors on the AIG-ASAP dataset. The results reveal that the existing detectors can be easily circumvented using straightforward automatic adversarial attacks. Specifically, we explore word substitution and sentence substitution perturbation methods that effectively evade detection while maintaining the quality of the generated essays. This highlights the urgent need for more accurate and robust methods to detect AI-generated student essays in the education domain. Code and data are released for public use.

Hidding the Ghostwriters: An Adversarial Evaluation of AI-Generated Student Essay Detection

emnlp-main.644

2402.00412

Oral

https://aclanthology.org/2023.emnlp-main.645.bib

https://aclanthology.org/2023.emnlp-main.645/

@inproceedings{wang-etal-2023-contextual, title = "Contextual Interaction for Argument Post Quality Assessment", author = "Wang, Yiran and Chen, Xuanang and He, Ben and Sun, Le", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.645", doi = "10.18653/v1/2023.emnlp-main.645", pages = "10420--10432", abstract = "Recently, there has been an increased emphasis on assessing the quality of natural language arguments. Existing approaches primarily focus on evaluating the quality of individual argument posts. However, they often fall short when it comes to effectively distinguishing arguments that possess a narrow quality margin. To address this limitation, this paper delves into two alternative methods for modeling the relative quality of different arguments. These approaches include: 1) Supervised contrastive learning that captures the intricate interactions between arguments. By incorporating this approach, we aim to enhance the assessment of argument quality by effectively distinguishing between arguments with subtle differences in quality. 2) Large language models (LLMs) with in-context examples that harness the power of LLMs and enrich them with in-context examples. Through extensive evaluation and analysis on the publicly available IBM-Rank-30k dataset, we demonstrate the superiority of our contrastive argument quality assessment approach over state-of-the-art baselines. On the other hand, while LLMs with in-context examples showcase a commendable ability to identify high-quality argument posts, they exhibit relatively limited efficacy in discerning between argument posts with a narrow quality gap.", }

Recently, there has been an increased emphasis on assessing the quality of natural language arguments. Existing approaches primarily focus on evaluating the quality of individual argument posts. However, they often fall short when it comes to effectively distinguishing arguments that possess a narrow quality margin. To address this limitation, this paper delves into two alternative methods for modeling the relative quality of different arguments. These approaches include: 1) Supervised contrastive learning that captures the intricate interactions between arguments. By incorporating this approach, we aim to enhance the assessment of argument quality by effectively distinguishing between arguments with subtle differences in quality. 2) Large language models (LLMs) with in-context examples that harness the power of LLMs and enrich them with in-context examples. Through extensive evaluation and analysis on the publicly available IBM-Rank-30k dataset, we demonstrate the superiority of our contrastive argument quality assessment approach over state-of-the-art baselines. On the other hand, while LLMs with in-context examples showcase a commendable ability to identify high-quality argument posts, they exhibit relatively limited efficacy in discerning between argument posts with a narrow quality gap.

Contextual Interaction for Argument Post Quality Assessment

emnlp-main.645

Poster

https://aclanthology.org/2023.emnlp-main.646.bib

https://aclanthology.org/2023.emnlp-main.646/

@inproceedings{sung-etal-2023-pre, title = "Pre-training Intent-Aware Encoders for Zero- and Few-Shot Intent Classification", author = "Sung, Mujeen and Gung, James and Mansimov, Elman and Pappas, Nikolaos and Shu, Raphael and Romeo, Salvatore and Zhang, Yi and Castelli, Vittorio", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.646", doi = "10.18653/v1/2023.emnlp-main.646", pages = "10433--10442", abstract = "Intent classification (IC) plays an important role in task-oriented dialogue systems. However, IC models often generalize poorly when training without sufficient annotated examples for each user intent. We propose a novel pre-training method for text encoders that uses contrastive learning with intent psuedo-labels to produce embeddings that are well-suited for IC tasks, reducing the need for manual annotations. By applying this pre-training strategy, we also introduce Pre-trained Intent-aware Encoder (PIE), which is designed to align encodings of utterances with their intent names. Specifically, we first train a tagger to identify key phrases within utterances that are crucial for interpreting intents. We then use these extracted phrases to create examples for pre-training a text encoder in a contrastive manner. As a result, our PIE model achieves up to 5.4{\%} and 4.0{\%} higher accuracy than the previous state-of-the-art pre-trained text encoder for the N-way zero- and one-shot settings on four IC datasets.", }

Intent classification (IC) plays an important role in task-oriented dialogue systems. However, IC models often generalize poorly when training without sufficient annotated examples for each user intent. We propose a novel pre-training method for text encoders that uses contrastive learning with intent psuedo-labels to produce embeddings that are well-suited for IC tasks, reducing the need for manual annotations. By applying this pre-training strategy, we also introduce Pre-trained Intent-aware Encoder (PIE), which is designed to align encodings of utterances with their intent names. Specifically, we first train a tagger to identify key phrases within utterances that are crucial for interpreting intents. We then use these extracted phrases to create examples for pre-training a text encoder in a contrastive manner. As a result, our PIE model achieves up to 5.4{\%} and 4.0{\%} higher accuracy than the previous state-of-the-art pre-trained text encoder for the N-way zero- and one-shot settings on four IC datasets.

Pre-training Intent-Aware Encoders for Zero- and Few-Shot Intent Classification

emnlp-main.646

2305.14827

Poster

https://aclanthology.org/2023.emnlp-main.647.bib

https://aclanthology.org/2023.emnlp-main.647/

@inproceedings{li-etal-2023-synthetic, title = "Synthetic Data Generation with Large Language Models for Text Classification: Potential and Limitations", author = "Li, Zhuoyan and Zhu, Hangxiao and Lu, Zhuoran and Yin, Ming", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.647", doi = "10.18653/v1/2023.emnlp-main.647", pages = "10443--10461", abstract = "The collection and curation of high-quality training data is crucial for developing text classification models with superior performance, but it is often associated with significant costs and time investment. Researchers have recently explored using large language models (LLMs) to generate synthetic datasets as an alternative approach. However, the effectiveness of the LLM-generated synthetic data in supporting model training is inconsistent across different classification tasks. To better understand factors that moderate the effectiveness of the LLM-generated synthetic data, in this study, we look into how the performance of models trained on these synthetic data may vary with the $\textit{subjectivity}$ of classification. Our results indicate that subjectivity, at both the task level and instance level, is negatively associated with the performance of the model trained on synthetic data. We conclude by discussing the implications of our work on the potential and limitations of leveraging LLM for synthetic data generation.", }

The collection and curation of high-quality training data is crucial for developing text classification models with superior performance, but it is often associated with significant costs and time investment. Researchers have recently explored using large language models (LLMs) to generate synthetic datasets as an alternative approach. However, the effectiveness of the LLM-generated synthetic data in supporting model training is inconsistent across different classification tasks. To better understand factors that moderate the effectiveness of the LLM-generated synthetic data, in this study, we look into how the performance of models trained on these synthetic data may vary with the $\textit{subjectivity}$ of classification. Our results indicate that subjectivity, at both the task level and instance level, is negatively associated with the performance of the model trained on synthetic data. We conclude by discussing the implications of our work on the potential and limitations of leveraging LLM for synthetic data generation.

Synthetic Data Generation with Large Language Models for Text Classification: Potential and Limitations

emnlp-main.647

2310.07849

https://huggingface.co/papers/2310.07849

Poster

https://aclanthology.org/2023.emnlp-main.648.bib

https://aclanthology.org/2023.emnlp-main.648/

@inproceedings{ilaslan-etal-2023-gazevqa, title = "{G}aze{VQA}: A Video Question Answering Dataset for Multiview Eye-Gaze Task-Oriented Collaborations", author = "Ilaslan, Muhammet and Song, Chenan and Chen, Joya and Gao, Difei and Lei, Weixian and Xu, Qianli and Lim, Joo and Shou, Mike", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.648", doi = "10.18653/v1/2023.emnlp-main.648", pages = "10462--10479", abstract = "The usage of exocentric and egocentric videos in Video Question Answering (VQA) is a new endeavor in human-robot interaction and collaboration studies. Particularly for egocentric videos, one may leverage eye-gaze information to understand human intentions during the task. In this paper, we build a novel task-oriented VQA dataset, called GazeVQA, for collaborative tasks where gaze information is captured during the task process. GazeVQA is designed with a novel QA format that covers thirteen different reasoning types to capture multiple aspects of task information and user intent. For each participant, GazeVQA consists of more than 1,100 textual questions and more than 500 labeled images that were annotated with the assistance of the Segment Anything Model. In total, 2,967 video clips, 12,491 labeled images, and 25,040 questions from 22 participants were included in the dataset. Additionally, inspired by the assisting models and common ground theory for industrial task collaboration, we propose a new AI model called AssistGaze that is designed to answer the questions with three different answer types, namely textual, image, and video. AssistGaze can effectively ground the perceptual input into semantic information while reducing ambiguities. We conduct comprehensive experiments to demonstrate the challenges of GazeVQA and the effectiveness of AssistGaze.", }

The usage of exocentric and egocentric videos in Video Question Answering (VQA) is a new endeavor in human-robot interaction and collaboration studies. Particularly for egocentric videos, one may leverage eye-gaze information to understand human intentions during the task. In this paper, we build a novel task-oriented VQA dataset, called GazeVQA, for collaborative tasks where gaze information is captured during the task process. GazeVQA is designed with a novel QA format that covers thirteen different reasoning types to capture multiple aspects of task information and user intent. For each participant, GazeVQA consists of more than 1,100 textual questions and more than 500 labeled images that were annotated with the assistance of the Segment Anything Model. In total, 2,967 video clips, 12,491 labeled images, and 25,040 questions from 22 participants were included in the dataset. Additionally, inspired by the assisting models and common ground theory for industrial task collaboration, we propose a new AI model called AssistGaze that is designed to answer the questions with three different answer types, namely textual, image, and video. AssistGaze can effectively ground the perceptual input into semantic information while reducing ambiguities. We conduct comprehensive experiments to demonstrate the challenges of GazeVQA and the effectiveness of AssistGaze.

GazeVQA: A Video Question Answering Dataset for Multiview Eye-Gaze Task-Oriented Collaborations

emnlp-main.648

Poster

https://aclanthology.org/2023.emnlp-main.649.bib

https://aclanthology.org/2023.emnlp-main.649/

@inproceedings{sen-etal-2023-people, title = "People Make Better Edits: Measuring the Efficacy of {LLM}-Generated Counterfactually Augmented Data for Harmful Language Detection", author = "Sen, Indira and Assenmacher, Dennis and Samory, Mattia and Augenstein, Isabelle and Aalst, Wil and Wagner, Claudia", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.649", doi = "10.18653/v1/2023.emnlp-main.649", pages = "10480--10504", abstract = "NLP models are used in a variety of critical social computing tasks, such as detecting sexist, racist, or otherwise hateful content. Therefore, it is imperative that these models are robust to spurious features. Past work has attempted to tackle such spurious features using training data augmentation, including Counterfactually Augmented Data (CADs). CADs introduce minimal changes to existing training data points and flip their labels; training on them may reduce model dependency on spurious features. However, manually generating CADs can be time-consuming and expensive. Hence in this work, we assess if this task can be automated using generative NLP models. We automatically generate CADs using Polyjuice, ChatGPT, and Flan-T5, and evaluate their usefulness in improving model robustness compared to manually-generated CADs. By testing both model performance on multiple out-of-domain test sets and individual data point efficacy, our results show that while manual CADs are still the most effective, CADs generated by ChatGPT come a close second. One key reason for the lower performance of automated methods is that the changes they introduce are often insufficient to flip the original label.", }

NLP models are used in a variety of critical social computing tasks, such as detecting sexist, racist, or otherwise hateful content. Therefore, it is imperative that these models are robust to spurious features. Past work has attempted to tackle such spurious features using training data augmentation, including Counterfactually Augmented Data (CADs). CADs introduce minimal changes to existing training data points and flip their labels; training on them may reduce model dependency on spurious features. However, manually generating CADs can be time-consuming and expensive. Hence in this work, we assess if this task can be automated using generative NLP models. We automatically generate CADs using Polyjuice, ChatGPT, and Flan-T5, and evaluate their usefulness in improving model robustness compared to manually-generated CADs. By testing both model performance on multiple out-of-domain test sets and individual data point efficacy, our results show that while manual CADs are still the most effective, CADs generated by ChatGPT come a close second. One key reason for the lower performance of automated methods is that the changes they introduce are often insufficient to flip the original label.

People Make Better Edits: Measuring the Efficacy of LLM-Generated Counterfactually Augmented Data for Harmful Language Detection

emnlp-main.649

2311.01270

https://huggingface.co/papers/2311.01270

Poster

https://aclanthology.org/2023.emnlp-main.650.bib

https://aclanthology.org/2023.emnlp-main.650/

@inproceedings{sun-etal-2023-unraveling, title = "Unraveling Feature Extraction Mechanisms in Neural Networks", author = "Sun, Xiaobing and Li, Jiaxi and Lu, Wei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.650", doi = "10.18653/v1/2023.emnlp-main.650", pages = "10505--10530", abstract = "The underlying mechanism of neural networks in capturing precise knowledge has been the subject of consistent research efforts. In this work, we propose a theoretical approach based on Neural Tangent Kernels (NTKs) to investigate such mechanisms. Specifically, considering the infinite network width, we hypothesize the learning dynamics of target models may intuitively unravel the features they acquire from training data, deepening our insights into their internal mechanisms. We apply our approach to several fundamental models and reveal how these models leverage statistical features during gradient descent and how they are integrated into final decisions. We also discovered that the choice of activation function can affect feature extraction. For instance, the use of the ReLU activation function could potentially introduce a bias in features, providing a plausible explanation for its replacement with alternative functions in recent pre-trained language models. Additionally, we find that while self-attention and CNN models may exhibit limitations in learning n-grams, multiplication-based models seem to excel in this area. We verify these theoretical findings through experiments and find that they can be applied to analyze language modeling tasks, which can be regarded as a special variant of classification. Our work may offer insights into the roles and capacities of fundamental modules within deep neural networks including large language models.", }

The underlying mechanism of neural networks in capturing precise knowledge has been the subject of consistent research efforts. In this work, we propose a theoretical approach based on Neural Tangent Kernels (NTKs) to investigate such mechanisms. Specifically, considering the infinite network width, we hypothesize the learning dynamics of target models may intuitively unravel the features they acquire from training data, deepening our insights into their internal mechanisms. We apply our approach to several fundamental models and reveal how these models leverage statistical features during gradient descent and how they are integrated into final decisions. We also discovered that the choice of activation function can affect feature extraction. For instance, the use of the ReLU activation function could potentially introduce a bias in features, providing a plausible explanation for its replacement with alternative functions in recent pre-trained language models. Additionally, we find that while self-attention and CNN models may exhibit limitations in learning n-grams, multiplication-based models seem to excel in this area. We verify these theoretical findings through experiments and find that they can be applied to analyze language modeling tasks, which can be regarded as a special variant of classification. Our work may offer insights into the roles and capacities of fundamental modules within deep neural networks including large language models.

Unraveling Feature Extraction Mechanisms in Neural Networks

emnlp-main.650

2310.16350

Poster

https://aclanthology.org/2023.emnlp-main.651.bib

https://aclanthology.org/2023.emnlp-main.651/

@inproceedings{he-etal-2023-capstone, title = "{CAPSTONE}: Curriculum Sampling for Dense Retrieval with Document Expansion", author = "He, Xingwei and Gong, Yeyun and Jin, A-Long and Zhang, Hang and Dong, Anlei and Jiao, Jian and Yiu, Siu and Duan, Nan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.651", doi = "10.18653/v1/2023.emnlp-main.651", pages = "10531--10541", abstract = "The dual-encoder has become the de facto architecture for dense retrieval. Typically, it computes the latent representations of the query and document independently, thus failing to fully capture the interactions between the query and document. To alleviate this, recent research has focused on obtaining query-informed document representations. During training, it expands the document with a real query, but during inference, it replaces the real query with a generated one. This inconsistency between training and inference causes the dense retrieval model to prioritize query information while disregarding the document when computing the document representation. Consequently, it performs even worse than the vanilla dense retrieval model because its performance heavily relies on the relevance between the generated queries and the real query. In this paper, we propose a curriculum sampling strategy that utilizes pseudo queries during training and progressively enhances the relevance between the generated query and the real query. By doing so, the retrieval model learns to extend its attention from the document alone to both the document and query, resulting in high-quality query-informed document representations. Experimental results on both in-domain and out-of-domain datasets demonstrate that our approach outperforms previous dense retrieval models.", }

The dual-encoder has become the de facto architecture for dense retrieval. Typically, it computes the latent representations of the query and document independently, thus failing to fully capture the interactions between the query and document. To alleviate this, recent research has focused on obtaining query-informed document representations. During training, it expands the document with a real query, but during inference, it replaces the real query with a generated one. This inconsistency between training and inference causes the dense retrieval model to prioritize query information while disregarding the document when computing the document representation. Consequently, it performs even worse than the vanilla dense retrieval model because its performance heavily relies on the relevance between the generated queries and the real query. In this paper, we propose a curriculum sampling strategy that utilizes pseudo queries during training and progressively enhances the relevance between the generated query and the real query. By doing so, the retrieval model learns to extend its attention from the document alone to both the document and query, resulting in high-quality query-informed document representations. Experimental results on both in-domain and out-of-domain datasets demonstrate that our approach outperforms previous dense retrieval models.

CAPSTONE: Curriculum Sampling for Dense Retrieval with Document Expansion

emnlp-main.651

Poster

https://aclanthology.org/2023.emnlp-main.652.bib

https://aclanthology.org/2023.emnlp-main.652/

@inproceedings{wang-etal-2023-balance, title = "Balance Act: Mitigating Hubness in Cross-Modal Retrieval with Query and Gallery Banks", author = "Wang, Yimu and Jian, Xiangru and Xue, Bo", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.652", doi = "10.18653/v1/2023.emnlp-main.652", pages = "10542--10567", abstract = "In this work, we present a post-processing solution to address the hubness problem in cross-modal retrieval, a phenomenon where a small number of gallery data points are frequently retrieved, resulting in a decline in retrieval performance. We first theoretically demonstrate the necessity of incorporating both the gallery and query data for addressing hubness as hubs always exhibit high similarity with gallery and query data. Second, building on our theoretical results, we propose a novel framework, Dual Bank Normalization (DBNorm). While previous work has attempted to alleviate hubness by only utilizing the query samples, DBNorm leverages two banks constructed from the query and gallery samples to reduce the occurrence of hubs during inference. Next, to complement DBNorm, we introduce two novel methods, dual inverted softmax and dual dynamic inverted softmax, for normalizing similarity based on the two banks. Specifically, our proposed methods reduce the similarity between hubs and queries while improving the similarity between non-hubs and queries. Finally, we present extensive experimental results on diverse language-grounded benchmarks, including text-image, text-video, and text-audio, demonstrating the superior performance of our approaches compared to previous methods in addressing hubness and boosting retrieval performance.", }

In this work, we present a post-processing solution to address the hubness problem in cross-modal retrieval, a phenomenon where a small number of gallery data points are frequently retrieved, resulting in a decline in retrieval performance. We first theoretically demonstrate the necessity of incorporating both the gallery and query data for addressing hubness as hubs always exhibit high similarity with gallery and query data. Second, building on our theoretical results, we propose a novel framework, Dual Bank Normalization (DBNorm). While previous work has attempted to alleviate hubness by only utilizing the query samples, DBNorm leverages two banks constructed from the query and gallery samples to reduce the occurrence of hubs during inference. Next, to complement DBNorm, we introduce two novel methods, dual inverted softmax and dual dynamic inverted softmax, for normalizing similarity based on the two banks. Specifically, our proposed methods reduce the similarity between hubs and queries while improving the similarity between non-hubs and queries. Finally, we present extensive experimental results on diverse language-grounded benchmarks, including text-image, text-video, and text-audio, demonstrating the superior performance of our approaches compared to previous methods in addressing hubness and boosting retrieval performance.

Balance Act: Mitigating Hubness in Cross-Modal Retrieval with Query and Gallery Banks

emnlp-main.652

2310.11612

https://huggingface.co/papers/2310.11612

Oral

https://aclanthology.org/2023.emnlp-main.653.bib

https://aclanthology.org/2023.emnlp-main.653/

@inproceedings{fu-etal-2023-e, title = "{E}-{CORE}: Emotion Correlation Enhanced Empathetic Dialogue Generation", author = "Fu, Fengyi and Zhang, Lei and Wang, Quan and Mao, Zhendong", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.653", doi = "10.18653/v1/2023.emnlp-main.653", pages = "10568--10586", abstract = "Achieving empathy is a crucial step toward humanized dialogue systems. Current approaches for empathetic dialogue generation mainly perceive an emotional label to generate an empathetic response conditioned on it, which simply treat emotions independently, but ignore the intrinsic emotion correlation in dialogues, resulting in inaccurate emotion perception and unsuitable response generation. In this paper, we propose a novel emotion correlation enhanced empathetic dialogue generation framework, which comprehensively realizes emotion correlation learning, utilization, and supervising. Specifically, a multi-resolution emotion graph is devised to capture context-based emotion interactions from different resolutions, further modeling emotion correlation. Then we propose an emotion correlation enhanced decoder, with a novel correlation-aware aggregation and soft/hard strategy, respectively improving the emotion perception and response generation. Experimental results on the benchmark dataset demonstrate the superiority of our model in both empathetic perception and expression.", }

Achieving empathy is a crucial step toward humanized dialogue systems. Current approaches for empathetic dialogue generation mainly perceive an emotional label to generate an empathetic response conditioned on it, which simply treat emotions independently, but ignore the intrinsic emotion correlation in dialogues, resulting in inaccurate emotion perception and unsuitable response generation. In this paper, we propose a novel emotion correlation enhanced empathetic dialogue generation framework, which comprehensively realizes emotion correlation learning, utilization, and supervising. Specifically, a multi-resolution emotion graph is devised to capture context-based emotion interactions from different resolutions, further modeling emotion correlation. Then we propose an emotion correlation enhanced decoder, with a novel correlation-aware aggregation and soft/hard strategy, respectively improving the emotion perception and response generation. Experimental results on the benchmark dataset demonstrate the superiority of our model in both empathetic perception and expression.

E-CORE: Emotion Correlation Enhanced Empathetic Dialogue Generation

emnlp-main.653

2311.15016

Poster

https://aclanthology.org/2023.emnlp-main.654.bib

https://aclanthology.org/2023.emnlp-main.654/

@inproceedings{gajbhiye-etal-2023-deck, title = "What do Deck Chairs and Sun Hats Have in Common? Uncovering Shared Properties in Large Concept Vocabularies", author = "Gajbhiye, Amit and Bouraoui, Zied and Li, Na and Chatterjee, Usashi and Espinosa-Anke, Luis and Schockaert, Steven", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.654", doi = "10.18653/v1/2023.emnlp-main.654", pages = "10587--10596", abstract = "Concepts play a central role in many applications. This includes settings where concepts have to be modelled in the absence of sentence context. Previous work has therefore focused on distilling decontextualised concept embeddings from language models. But concepts can be modelled from different perspectives, whereas concept embeddings typically mostly capture taxonomic structure. To address this issue, we propose a strategy for identifying what different concepts, from a potentially large concept vocabulary, have in common with others. We then represent concepts in terms of the properties they share with the other concepts. To demonstrate the practical usefulness of this way of modelling concepts, we consider the task of ultra-fine entity typing, which is a challenging multi-label classification problem. We show that by augmenting the label set with shared properties, we can improve the performance of the state-of-the-art models for this task.", }

Concepts play a central role in many applications. This includes settings where concepts have to be modelled in the absence of sentence context. Previous work has therefore focused on distilling decontextualised concept embeddings from language models. But concepts can be modelled from different perspectives, whereas concept embeddings typically mostly capture taxonomic structure. To address this issue, we propose a strategy for identifying what different concepts, from a potentially large concept vocabulary, have in common with others. We then represent concepts in terms of the properties they share with the other concepts. To demonstrate the practical usefulness of this way of modelling concepts, we consider the task of ultra-fine entity typing, which is a challenging multi-label classification problem. We show that by augmenting the label set with shared properties, we can improve the performance of the state-of-the-art models for this task.

What do Deck Chairs and Sun Hats Have in Common? Uncovering Shared Properties in Large Concept Vocabularies

emnlp-main.654

2310.14793

Oral

https://aclanthology.org/2023.emnlp-main.655.bib

https://aclanthology.org/2023.emnlp-main.655/

@inproceedings{keleg-etal-2023-aldi, title = "{ALD}i: Quantifying the {A}rabic Level of Dialectness of Text", author = "Keleg, Amr and Goldwater, Sharon and Magdy, Walid", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.655", doi = "10.18653/v1/2023.emnlp-main.655", pages = "10597--10611", abstract = "Transcribed speech and user-generated text in Arabic typically contain a mixture of Modern Standard Arabic (MSA), the standardized language taught in schools, and Dialectal Arabic (DA), used in daily communications. To handle this variation, previous work in Arabic NLP has focused on Dialect Identification (DI) on the sentence or the token level. However, DI treats the task as binary, whereas we argue that Arabic speakers perceive a spectrum of dialectness, which we operationalize at the sentence level as the Arabic Level of Dialectness (ALDi), a continuous linguistic variable. We introduce the AOC-ALDi dataset (derived from the AOC dataset), containing 127,835 sentences (17{\%} from news articles and 83{\%} from user comments on those articles) which are manually labeled with their level of dialectness. We provide a detailed analysis of AOC-ALDi and show that a model trained on it can effectively identify levels of dialectness on a range of other corpora (including dialects and genres not included in AOC-ALDi), providing a more nuanced picture than traditional DI systems. Through case studies, we illustrate how ALDi can reveal Arabic speakers{'} stylistic choices in different situations, a useful property for sociolinguistic analyses.", }

Transcribed speech and user-generated text in Arabic typically contain a mixture of Modern Standard Arabic (MSA), the standardized language taught in schools, and Dialectal Arabic (DA), used in daily communications. To handle this variation, previous work in Arabic NLP has focused on Dialect Identification (DI) on the sentence or the token level. However, DI treats the task as binary, whereas we argue that Arabic speakers perceive a spectrum of dialectness, which we operationalize at the sentence level as the Arabic Level of Dialectness (ALDi), a continuous linguistic variable. We introduce the AOC-ALDi dataset (derived from the AOC dataset), containing 127,835 sentences (17{\%} from news articles and 83{\%} from user comments on those articles) which are manually labeled with their level of dialectness. We provide a detailed analysis of AOC-ALDi and show that a model trained on it can effectively identify levels of dialectness on a range of other corpora (including dialects and genres not included in AOC-ALDi), providing a more nuanced picture than traditional DI systems. Through case studies, we illustrate how ALDi can reveal Arabic speakers{'} stylistic choices in different situations, a useful property for sociolinguistic analyses.

ALDi: Quantifying the Arabic Level of Dialectness of Text

emnlp-main.655

Poster

https://aclanthology.org/2023.emnlp-main.656.bib

https://aclanthology.org/2023.emnlp-main.656/

@inproceedings{wang-etal-2023-3drp, title = "3{DRP}-Net: 3{D} Relative Position-aware Network for 3{D} Visual Grounding", author = "Wang, Zehan and Huang, Haifeng and Zhao, Yang and Li, Linjun and Cheng, Xize and Zhu, Yichen and Yin, Aoxiong and Zhao, Zhou", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.656", doi = "10.18653/v1/2023.emnlp-main.656", pages = "10612--10625", abstract = "3D visual grounding aims to localize the target object in a 3D point cloud by a free-form language description. Typically, the sentences describing the target object tend to provide information about its relative relation between other objects and its position within the whole scene. In this work, we propose a relation-aware one-stage framework, named 3D Relative Position-aware Network (3DRP-Net), which can effectively capture the relative spatial relationships between objects and enhance object attributes. Specifically, 1) we propose a 3D Relative Position Multi-head Attention (3DRP-MA) module to analyze relative relations from different directions in the context of object pairs, which helps the model to focus on the specific object relations mentioned in the sentence. 2) We designed a soft-labeling strategy to alleviate the spatial ambiguity caused by redundant points, which further stabilizes and enhances the learning process through a constant and discriminative distribution. Extensive experiments conducted on three benchmarks (i.e., ScanRefer and Nr3D/Sr3D) demonstrate that our method outperforms all the state-of-the-art methods in general.", }

3D visual grounding aims to localize the target object in a 3D point cloud by a free-form language description. Typically, the sentences describing the target object tend to provide information about its relative relation between other objects and its position within the whole scene. In this work, we propose a relation-aware one-stage framework, named 3D Relative Position-aware Network (3DRP-Net), which can effectively capture the relative spatial relationships between objects and enhance object attributes. Specifically, 1) we propose a 3D Relative Position Multi-head Attention (3DRP-MA) module to analyze relative relations from different directions in the context of object pairs, which helps the model to focus on the specific object relations mentioned in the sentence. 2) We designed a soft-labeling strategy to alleviate the spatial ambiguity caused by redundant points, which further stabilizes and enhances the learning process through a constant and discriminative distribution. Extensive experiments conducted on three benchmarks (i.e., ScanRefer and Nr3D/Sr3D) demonstrate that our method outperforms all the state-of-the-art methods in general.

3DRP-Net: 3D Relative Position-aware Network for 3D Visual Grounding

emnlp-main.656

2307.13363

Oral

https://aclanthology.org/2023.emnlp-main.657.bib

https://aclanthology.org/2023.emnlp-main.657/

@inproceedings{wang-etal-2023-goal, title = "Goal-Driven Explainable Clustering via Language Descriptions", author = "Wang, Zihan and Shang, Jingbo and Zhong, Ruiqi", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.657", doi = "10.18653/v1/2023.emnlp-main.657", pages = "10626--10649", abstract = "Unsupervised clustering is widely used to explore large corpora, but existing formulations neither consider the users{'} goals nor explain clusters{'} meanings. We propose a new task formulation, {``}Goal-Driven Clustering with Explanations{''} (GoalEx), which represents both the goal and the explanations as free-form language descriptions. For example, to categorize the errors made by a summarization system, the input to GoalEx is a corpus of annotator-written comments for system-generated summaries and a goal description {``}cluster the comments based on why the annotators think the summary is imperfect.{''}; the outputs are text clusters each with an explanation ({``}this cluster mentions that the summary misses important context information.{''}), which relates to the goal and accurately explains which comments should (not) belong to a cluster. To tackle GoalEx, we prompt a language model with {``}[corpus subset] + [goal] + Brainstorm a list of explanations each representing a cluster.{''}; then we classify whether each sample belongs to a cluster based on its explanation; finally, we use integer linear programming to select a subset of candidate clusters to cover most samples while minimizing overlaps. Under both automatic and human evaluation on corpora with or without labels, our method produces more accurate and goal-related explanations than prior methods.", }

Unsupervised clustering is widely used to explore large corpora, but existing formulations neither consider the users{'} goals nor explain clusters{'} meanings. We propose a new task formulation, {``}Goal-Driven Clustering with Explanations{''} (GoalEx), which represents both the goal and the explanations as free-form language descriptions. For example, to categorize the errors made by a summarization system, the input to GoalEx is a corpus of annotator-written comments for system-generated summaries and a goal description {``}cluster the comments based on why the annotators think the summary is imperfect.{''}; the outputs are text clusters each with an explanation ({``}this cluster mentions that the summary misses important context information.{''}), which relates to the goal and accurately explains which comments should (not) belong to a cluster. To tackle GoalEx, we prompt a language model with {``}[corpus subset] + [goal] + Brainstorm a list of explanations each representing a cluster.{''}; then we classify whether each sample belongs to a cluster based on its explanation; finally, we use integer linear programming to select a subset of candidate clusters to cover most samples while minimizing overlaps. Under both automatic and human evaluation on corpora with or without labels, our method produces more accurate and goal-related explanations than prior methods.

Goal-Driven Explainable Clustering via Language Descriptions

emnlp-main.657

2305.13749

https://huggingface.co/papers/2305.13749

Oral

https://aclanthology.org/2023.emnlp-main.658.bib

https://aclanthology.org/2023.emnlp-main.658/

@inproceedings{qi-etal-2023-cross, title = "Cross-Lingual Consistency of Factual Knowledge in Multilingual Language Models", author = "Qi, Jirui and Fern{\'a}ndez, Raquel and Bisazza, Arianna", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.658", doi = "10.18653/v1/2023.emnlp-main.658", pages = "10650--10666", abstract = "Multilingual large-scale Pretrained Language Models (PLMs) have been shown to store considerable amounts of factual knowledge, but large variations are observed across languages. With the ultimate goal of ensuring that users with different language backgrounds obtain consistent feedback from the same model, we study the cross-lingual consistency (CLC) of factual knowledge in various multilingual PLMs. To this end, we propose a Ranking-based Consistency (RankC) metric to evaluate knowledge consistency across languages independently from accuracy. Using this metric, we conduct an in-depth analysis of the determining factors for CLC, both at model level and at language-pair level. Among other results, we find that increasing model size leads to higher factual probing accuracy in most languages, but does not improve cross-lingual consistency. Finally, we conduct a case study on CLC when new factual associations are inserted in the PLMs via model editing. Results on a small sample of facts inserted in English reveal a clear pattern whereby the new piece of knowledge transfers only to languages with which English has a high RankC score. All code and data are released at https://github.com/Betswish/Cross-Lingual-Consistency.", }

Multilingual large-scale Pretrained Language Models (PLMs) have been shown to store considerable amounts of factual knowledge, but large variations are observed across languages. With the ultimate goal of ensuring that users with different language backgrounds obtain consistent feedback from the same model, we study the cross-lingual consistency (CLC) of factual knowledge in various multilingual PLMs. To this end, we propose a Ranking-based Consistency (RankC) metric to evaluate knowledge consistency across languages independently from accuracy. Using this metric, we conduct an in-depth analysis of the determining factors for CLC, both at model level and at language-pair level. Among other results, we find that increasing model size leads to higher factual probing accuracy in most languages, but does not improve cross-lingual consistency. Finally, we conduct a case study on CLC when new factual associations are inserted in the PLMs via model editing. Results on a small sample of facts inserted in English reveal a clear pattern whereby the new piece of knowledge transfers only to languages with which English has a high RankC score. All code and data are released at https://github.com/Betswish/Cross-Lingual-Consistency.

Cross-Lingual Consistency of Factual Knowledge in Multilingual Language Models

emnlp-main.658

Poster

https://aclanthology.org/2023.emnlp-main.659.bib

https://aclanthology.org/2023.emnlp-main.659/

@inproceedings{wang-li-2023-learning, title = "Learning from Mistakes via Cooperative Study Assistant for Large Language Models", author = "Wang, Danqing and Li, Lei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.659", doi = "10.18653/v1/2023.emnlp-main.659", pages = "10667--10685", abstract = "Large language models (LLMs) have demonstrated their potential to refine their generation based on their own feedback. However, the feedback from LLM itself is often inaccurate, thereby limiting its benefits. In this paper, we propose Study Assistant for Large LAnguage Model (SALAM), a novel framework with an auxiliary agent to assist the main LLM in learning from mistakes through interactive cooperation. In the gathering phase, the student assistant agent probes the main LLM, analyzes its errors, and collects the interaction in a mistake memory. During the examination phase, the study assistant provides guidelines by retrieving relevant cases to help the main LLM anticipate and avoid similar errors. We first investigate the effectiveness of a general study assistant and then customize it to provide LLM-specific guidance through imitation learning from successful guidance experiences. Our experiments on three LLMs using two challenging frameworks demonstrate that SALAM can significantly boost LLMs by an accuracy margin of up to 6.6 on BBH and 12.6 on BBQ.", }

Large language models (LLMs) have demonstrated their potential to refine their generation based on their own feedback. However, the feedback from LLM itself is often inaccurate, thereby limiting its benefits. In this paper, we propose Study Assistant for Large LAnguage Model (SALAM), a novel framework with an auxiliary agent to assist the main LLM in learning from mistakes through interactive cooperation. In the gathering phase, the student assistant agent probes the main LLM, analyzes its errors, and collects the interaction in a mistake memory. During the examination phase, the study assistant provides guidelines by retrieving relevant cases to help the main LLM anticipate and avoid similar errors. We first investigate the effectiveness of a general study assistant and then customize it to provide LLM-specific guidance through imitation learning from successful guidance experiences. Our experiments on three LLMs using two challenging frameworks demonstrate that SALAM can significantly boost LLMs by an accuracy margin of up to 6.6 on BBH and 12.6 on BBQ.

Learning from Mistakes via Cooperative Study Assistant for Large Language Models

emnlp-main.659

2305.13829

Poster

https://aclanthology.org/2023.emnlp-main.660.bib

https://aclanthology.org/2023.emnlp-main.660/

@inproceedings{nwatu-etal-2023-bridging, title = "Bridging the Digital Divide: Performance Variation across Socio-Economic Factors in Vision-Language Models", author = "Nwatu, Joan and Ignat, Oana and Mihalcea, Rada", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.660", doi = "10.18653/v1/2023.emnlp-main.660", pages = "10686--10702", abstract = "Despite the impressive performance of current AI models reported across various tasks, performance reports often do not include evaluations of how these models perform on the specific groups that will be impacted by these technologies. Among the minority groups under-represented in AI, data from low-income households are often overlooked in data collection and model evaluation. We evaluate the performance of a state-of-the-art vision-language model (CLIP) on a geo-diverse dataset containing household images associated with different income values (DollarStreet) and show that performance inequality exists among households of different income levels. Our results indicate that performance for the poorer groups is consistently lower than the wealthier groups across various topics and countries. We highlight insights that can help mitigate these issues and propose actionable steps for economic-level inclusive AI development.", }

Despite the impressive performance of current AI models reported across various tasks, performance reports often do not include evaluations of how these models perform on the specific groups that will be impacted by these technologies. Among the minority groups under-represented in AI, data from low-income households are often overlooked in data collection and model evaluation. We evaluate the performance of a state-of-the-art vision-language model (CLIP) on a geo-diverse dataset containing household images associated with different income values (DollarStreet) and show that performance inequality exists among households of different income levels. Our results indicate that performance for the poorer groups is consistently lower than the wealthier groups across various topics and countries. We highlight insights that can help mitigate these issues and propose actionable steps for economic-level inclusive AI development.

Bridging the Digital Divide: Performance Variation across Socio-Economic Factors in Vision-Language Models

emnlp-main.660

2311.05746

Oral

https://aclanthology.org/2023.emnlp-main.661.bib

https://aclanthology.org/2023.emnlp-main.661/

@inproceedings{yifei-etal-2023-conceptor, title = "Conceptor-Aided Debiasing of Large Language Models", author = "Yifei, Li and Ungar, Lyle and Sedoc, Jo{\~a}o", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.661", doi = "10.18653/v1/2023.emnlp-main.661", pages = "10703--10727", abstract = "Pre-trained large language models (LLMs) reflect the inherent social biases of their training corpus. Many methods have been proposed to mitigate this issue, but they often fail to debias or they sacrifice model accuracy. We use *conceptors*{--}a soft projection method{--}to identify and remove the bias subspace in LLMs such as BERT and GPT. We propose two methods of applying conceptors (1) bias subspace projection by post-processing by the conceptor NOT operation; and (2) a new architecture, conceptor-intervened BERT (CI-BERT), which explicitly incorporates the conceptor projection into all layers during training. We find that conceptor post-processing achieves state-of-the-art (SoTA) debiasing results while maintaining LLMs{'} performance on the GLUE benchmark. Further, it is robust in various scenarios and can mitigate intersectional bias efficiently by its AND operation on the existing bias subspaces. Although CI-BERT{'}s training takes all layers{'} bias into account and can beat its post-processing counterpart in bias mitigation, CI-BERT reduces the language model accuracy. We also show the importance of carefully constructing the bias subspace. The best results are obtained by removing outliers from the list of biased words, combining them (via the OR operation), and computing their embeddings using the sentences from a cleaner corpus.", }

Pre-trained large language models (LLMs) reflect the inherent social biases of their training corpus. Many methods have been proposed to mitigate this issue, but they often fail to debias or they sacrifice model accuracy. We use *conceptors*{--}a soft projection method{--}to identify and remove the bias subspace in LLMs such as BERT and GPT. We propose two methods of applying conceptors (1) bias subspace projection by post-processing by the conceptor NOT operation; and (2) a new architecture, conceptor-intervened BERT (CI-BERT), which explicitly incorporates the conceptor projection into all layers during training. We find that conceptor post-processing achieves state-of-the-art (SoTA) debiasing results while maintaining LLMs{'} performance on the GLUE benchmark. Further, it is robust in various scenarios and can mitigate intersectional bias efficiently by its AND operation on the existing bias subspaces. Although CI-BERT{'}s training takes all layers{'} bias into account and can beat its post-processing counterpart in bias mitigation, CI-BERT reduces the language model accuracy. We also show the importance of carefully constructing the bias subspace. The best results are obtained by removing outliers from the list of biased words, combining them (via the OR operation), and computing their embeddings using the sentences from a cleaner corpus.

Conceptor-Aided Debiasing of Large Language Models

emnlp-main.661

Poster

https://aclanthology.org/2023.emnlp-main.662.bib

https://aclanthology.org/2023.emnlp-main.662/

@inproceedings{groschwitz-etal-2023-amr, title = "{AMR} Parsing is Far from Solved: {G}r{APES}, the Granular {AMR} Parsing Evaluation Suite", author = "Groschwitz, Jonas and Cohen, Shay and Donatelli, Lucia and Fowlie, Meaghan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.662", doi = "10.18653/v1/2023.emnlp-main.662", pages = "10728--10752", abstract = "We present the Granular AMR Parsing Evaluation Suite (GrAPES), a challenge set for Abstract Meaning Representation (AMR) parsing with accompanying evaluation metrics. AMR parsers now obtain high scores on the standard AMR evaluation metric Smatch, close to or even above reported inter-annotator agreement. But that does not mean that AMR parsing is solved; in fact, human evaluation in previous work indicates that current parsers still quite frequently make errors on node labels or graph structure that substantially distort sentence meaning. Here, we provide an evaluation suite that tests AMR parsers on a range of phenomena of practical, technical, and linguistic interest. Our 36 categories range from seen and unseen labels, to structural generalization, to coreference. GrAPES reveals in depth the abilities and shortcomings of current AMR parsers.", }

We present the Granular AMR Parsing Evaluation Suite (GrAPES), a challenge set for Abstract Meaning Representation (AMR) parsing with accompanying evaluation metrics. AMR parsers now obtain high scores on the standard AMR evaluation metric Smatch, close to or even above reported inter-annotator agreement. But that does not mean that AMR parsing is solved; in fact, human evaluation in previous work indicates that current parsers still quite frequently make errors on node labels or graph structure that substantially distort sentence meaning. Here, we provide an evaluation suite that tests AMR parsers on a range of phenomena of practical, technical, and linguistic interest. Our 36 categories range from seen and unseen labels, to structural generalization, to coreference. GrAPES reveals in depth the abilities and shortcomings of current AMR parsers.

AMR Parsing is Far from Solved: GrAPES, the Granular AMR Parsing Evaluation Suite

emnlp-main.662

2312.03480

Poster

https://aclanthology.org/2023.emnlp-main.663.bib

https://aclanthology.org/2023.emnlp-main.663/

@inproceedings{zhang-etal-2023-rethinking, title = "Rethinking and Improving Multi-task Learning for End-to-end Speech Translation", author = "Zhang, Yuhao and Xu, Chen and Li, Bei and Chen, Hao and Xiao, Tong and Zhang, Chunliang and Zhu, Jingbo", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.663", doi = "10.18653/v1/2023.emnlp-main.663", pages = "10753--10765", abstract = "Significant improvements in end-to-end speech translation (ST) have been achieved through the application of multi-task learning. However, the extent to which auxiliary tasks are highly consistent with the ST task, and how much this approach truly helps, have not been thoroughly studied. In this paper, we investigate the consistency between different tasks, considering different times and modules. We find that the textual encoder primarily facilitates cross-modal conversion, but the presence of noise in speech impedes the consistency between text and speech representations. Furthermore, we propose an improved multi-task learning (IMTL) approach for the ST task, which bridges the modal gap by mitigating the difference in length and representation. We conduct experiments on the MuST-C dataset. The results demonstrate that our method attains state-of-the-art results. Moreover, when additional data is used, we achieve the new SOTA result on MuST-C English to Spanish task with 20.8{\%} of the training time required by the current SOTA method.", }

Significant improvements in end-to-end speech translation (ST) have been achieved through the application of multi-task learning. However, the extent to which auxiliary tasks are highly consistent with the ST task, and how much this approach truly helps, have not been thoroughly studied. In this paper, we investigate the consistency between different tasks, considering different times and modules. We find that the textual encoder primarily facilitates cross-modal conversion, but the presence of noise in speech impedes the consistency between text and speech representations. Furthermore, we propose an improved multi-task learning (IMTL) approach for the ST task, which bridges the modal gap by mitigating the difference in length and representation. We conduct experiments on the MuST-C dataset. The results demonstrate that our method attains state-of-the-art results. Moreover, when additional data is used, we achieve the new SOTA result on MuST-C English to Spanish task with 20.8{\%} of the training time required by the current SOTA method.

Rethinking and Improving Multi-task Learning for End-to-end Speech Translation

emnlp-main.663

2311.03810

Oral

https://aclanthology.org/2023.emnlp-main.664.bib

https://aclanthology.org/2023.emnlp-main.664/

@inproceedings{bejan-etal-2023-ad, title = "{AD}-{NLP}: A Benchmark for Anomaly Detection in Natural Language Processing", author = "Bejan, Matei and Manolache, Andrei and Popescu, Marius", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.664", doi = "10.18653/v1/2023.emnlp-main.664", pages = "10766--10778", abstract = "Deep learning models have reignited the interest in Anomaly Detection research in recent years. Methods for Anomaly Detection in text have shown strong empirical results on ad-hoc anomaly setups that are usually made by downsampling some classes of a labeled dataset. This can lead to reproducibility issues and models that are biased toward detecting particular anomalies while failing to recognize them in more sophisticated scenarios. In the present work, we provide a unified benchmark for detecting various types of anomalies, focusing on problems that can be naturally formulated as Anomaly Detection in text, ranging from syntax to stylistics. In this way, we are hoping to facilitate research in Text Anomaly Detection. We also evaluate and analyze two strong shallow baselines, as well as two of the current state-of-the-art neural approaches, providing insights into the knowledge the neural models are learning when performing the anomaly detection task. We provide the code for evaluation, downloading, and preprocessing the dataset at https://github.com/mateibejan1/ad-nlp/.", }

Deep learning models have reignited the interest in Anomaly Detection research in recent years. Methods for Anomaly Detection in text have shown strong empirical results on ad-hoc anomaly setups that are usually made by downsampling some classes of a labeled dataset. This can lead to reproducibility issues and models that are biased toward detecting particular anomalies while failing to recognize them in more sophisticated scenarios. In the present work, we provide a unified benchmark for detecting various types of anomalies, focusing on problems that can be naturally formulated as Anomaly Detection in text, ranging from syntax to stylistics. In this way, we are hoping to facilitate research in Text Anomaly Detection. We also evaluate and analyze two strong shallow baselines, as well as two of the current state-of-the-art neural approaches, providing insights into the knowledge the neural models are learning when performing the anomaly detection task. We provide the code for evaluation, downloading, and preprocessing the dataset at https://github.com/mateibejan1/ad-nlp/.

AD-NLP: A Benchmark for Anomaly Detection in Natural Language Processing

emnlp-main.664

Poster

https://aclanthology.org/2023.emnlp-main.665.bib

https://aclanthology.org/2023.emnlp-main.665/

@inproceedings{zerveas-etal-2023-enhancing, title = "Enhancing the Ranking Context of Dense Retrieval through Reciprocal Nearest Neighbors", author = "Zerveas, George and Rekabsaz, Navid and Eickhoff, Carsten", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.665", doi = "10.18653/v1/2023.emnlp-main.665", pages = "10779--10803", abstract = "Sparse annotation poses persistent challenges to training dense retrieval models; for example, it distorts the training signal when unlabeled relevant documents are used spuriously as negatives in contrastive learning. To alleviate this problem, we introduce evidence-based label smoothing, a novel, computationally efficient method that prevents penalizing the model for assigning high relevance to false negatives. To compute the target relevance distribution over candidate documents within the ranking context of a given query, we assign a non-zero relevance probability to those candidates most similar to the ground truth based on the degree of their similarity to the ground-truth document(s). To estimate relevance we leverage an improved similarity metric based on reciprocal nearest neighbors, which can also be used independently to rerank candidates in post-processing. Through extensive experiments on two large-scale ad hoc text retrieval datasets, we demonstrate that reciprocal nearest neighbors can improve the ranking effectiveness of dense retrieval models, both when used for label smoothing, as well as for reranking. This indicates that by considering relationships between documents and queries beyond simple geometric distance we can effectively enhance the ranking context.", }

Sparse annotation poses persistent challenges to training dense retrieval models; for example, it distorts the training signal when unlabeled relevant documents are used spuriously as negatives in contrastive learning. To alleviate this problem, we introduce evidence-based label smoothing, a novel, computationally efficient method that prevents penalizing the model for assigning high relevance to false negatives. To compute the target relevance distribution over candidate documents within the ranking context of a given query, we assign a non-zero relevance probability to those candidates most similar to the ground truth based on the degree of their similarity to the ground-truth document(s). To estimate relevance we leverage an improved similarity metric based on reciprocal nearest neighbors, which can also be used independently to rerank candidates in post-processing. Through extensive experiments on two large-scale ad hoc text retrieval datasets, we demonstrate that reciprocal nearest neighbors can improve the ranking effectiveness of dense retrieval models, both when used for label smoothing, as well as for reranking. This indicates that by considering relationships between documents and queries beyond simple geometric distance we can effectively enhance the ranking context.

Enhancing the Ranking Context of Dense Retrieval through Reciprocal Nearest Neighbors

emnlp-main.665

Poster

https://aclanthology.org/2023.emnlp-main.666.bib

https://aclanthology.org/2023.emnlp-main.666/

@inproceedings{zhang-etal-2023-cross-lingual, title = "Cross-Lingual Cross-Target Stance Detection with Dual Knowledge Distillation Framework", author = "Zhang, Ruike and Yang, Hanxuan and Mao, Wenji", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.666", doi = "10.18653/v1/2023.emnlp-main.666", pages = "10804--10819", abstract = "Stance detection aims to identify the user{'}s attitude toward specific \textit{targets} from text, which is an important research area in text mining and benefits a variety of application domains. Existing studies on stance detection were conducted mainly in English. Due to the low-resource problem in most non-English languages, cross-lingual stance detection was proposed to transfer knowledge from high-resource (source) language to low-resource (target) language. However, previous research has ignored the practical issue of no labeled training data available in target language. Moreover, target inconsistency in cross-lingual stance detection brings about the additional issue of unseen targets in target language, which in essence requires the transfer of both language and target-oriented knowledge from source to target language. To tackle these challenging issues, in this paper, we propose the new task of cross-lingual cross-target stance detection and develop the first computational work with dual knowledge distillation. Our proposed framework designs a cross-lingual teacher and a cross-target teacher using the source language data and a dual distillation process that transfers the two types of knowledge to target language. To bridge the target discrepancy between languages, cross-target teacher mines target category information and generalizes it to the unseen targets in target language via category-oriented learning. Experimental results on multilingual stance datasets demonstrate the effectiveness of our method compared to the competitive baselines.", }

Stance detection aims to identify the user{'}s attitude toward specific \textit{targets} from text, which is an important research area in text mining and benefits a variety of application domains. Existing studies on stance detection were conducted mainly in English. Due to the low-resource problem in most non-English languages, cross-lingual stance detection was proposed to transfer knowledge from high-resource (source) language to low-resource (target) language. However, previous research has ignored the practical issue of no labeled training data available in target language. Moreover, target inconsistency in cross-lingual stance detection brings about the additional issue of unseen targets in target language, which in essence requires the transfer of both language and target-oriented knowledge from source to target language. To tackle these challenging issues, in this paper, we propose the new task of cross-lingual cross-target stance detection and develop the first computational work with dual knowledge distillation. Our proposed framework designs a cross-lingual teacher and a cross-target teacher using the source language data and a dual distillation process that transfers the two types of knowledge to target language. To bridge the target discrepancy between languages, cross-target teacher mines target category information and generalizes it to the unseen targets in target language via category-oriented learning. Experimental results on multilingual stance datasets demonstrate the effectiveness of our method compared to the competitive baselines.

Cross-Lingual Cross-Target Stance Detection with Dual Knowledge Distillation Framework

emnlp-main.666

Poster

https://aclanthology.org/2023.emnlp-main.667.bib

https://aclanthology.org/2023.emnlp-main.667/

@inproceedings{goel-etal-2023-presto, title = "{PRESTO}: A Multilingual Dataset for Parsing Realistic Task-Oriented Dialogs", author = "Goel, Rahul and Ammar, Waleed and Gupta, Aditya and Vashishtha, Siddharth and Sano, Motoki and Surani, Faiz and Chang, Max and Choe, HyunJeong and Greene, David and He, Chuan and Nitisaroj, Rattima and Trukhina, Anna and Paul, Shachi and Shah, Pararth and Shah, Rushin and Yu, Zhou", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.667", doi = "10.18653/v1/2023.emnlp-main.667", pages = "10820--10833", abstract = "Research interest in task-oriented dialogs has increased as systems such as Google Assistant, Alexa and Siri have become ubiquitous in everyday life. However, the impact of academic research in this area has been limited by the lack of datasets that realistically capture the wide array of user pain points. To enable research on some of the more challenging aspects of parsing realistic conversations, we introduce PRESTO, a public dataset of over 550K contextual multilingual conversations between humans and virtual assistants. PRESTO contains a diverse array of challenges that occur in real-world NLU tasks such as disfluencies, code-switching, and revisions. It is the only large scale human generated conversational parsing dataset that provides structured context such as a user{'}s contacts and lists for each example. Our mT5 model based baselines demonstrate that the conversational phenomenon present in PRESTO are challenging to model, which is further pronounced in a low-resource setup.", }

Research interest in task-oriented dialogs has increased as systems such as Google Assistant, Alexa and Siri have become ubiquitous in everyday life. However, the impact of academic research in this area has been limited by the lack of datasets that realistically capture the wide array of user pain points. To enable research on some of the more challenging aspects of parsing realistic conversations, we introduce PRESTO, a public dataset of over 550K contextual multilingual conversations between humans and virtual assistants. PRESTO contains a diverse array of challenges that occur in real-world NLU tasks such as disfluencies, code-switching, and revisions. It is the only large scale human generated conversational parsing dataset that provides structured context such as a user{'}s contacts and lists for each example. Our mT5 model based baselines demonstrate that the conversational phenomenon present in PRESTO are challenging to model, which is further pronounced in a low-resource setup.

PRESTO: A Multilingual Dataset for Parsing Realistic Task-Oriented Dialogs

emnlp-main.667

2303.08954

https://huggingface.co/papers/2303.08954

Poster

https://aclanthology.org/2023.emnlp-main.668.bib

https://aclanthology.org/2023.emnlp-main.668/

@inproceedings{huang-etal-2023-iteratively, title = "An Iteratively Parallel Generation Method with the Pre-Filling Strategy for Document-level Event Extraction", author = "Huang, Guanhua and Xu, Runxin and Zeng, Ying and Chen, Jiaze and Yang, Zhouwang and E, Weinan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.668", doi = "10.18653/v1/2023.emnlp-main.668", pages = "10834--10852", abstract = "In document-level event extraction (DEE) tasks, a document typically contains many event records with multiple event roles. Therefore, accurately extracting all event records is a big challenge since the number of event records is not given. Previous works present the entity-based directed acyclic graph (EDAG) generation methods to autoregressively generate event roles, which requires a given generation order. Meanwhile, parallel methods are proposed to generate all event roles simultaneously, but suffer from the inadequate training which manifests zero accuracies on some event roles. In this paper, we propose an Iteratively Parallel Generation method with the Pre-Filling strategy (IPGPF). Event roles in an event record are generated in parallel to avoid order selection, and the event records are iteratively generated to utilize historical results. Experiments on two public datasets show our IPGPF improves 11.7 F1 than previous parallel models and up to 5.1 F1 than auto-regressive models under the control variable settings. Moreover, our enhanced IPGPF outperforms other entity-enhanced models and achieves new state-of-the-art performance.", }

In document-level event extraction (DEE) tasks, a document typically contains many event records with multiple event roles. Therefore, accurately extracting all event records is a big challenge since the number of event records is not given. Previous works present the entity-based directed acyclic graph (EDAG) generation methods to autoregressively generate event roles, which requires a given generation order. Meanwhile, parallel methods are proposed to generate all event roles simultaneously, but suffer from the inadequate training which manifests zero accuracies on some event roles. In this paper, we propose an Iteratively Parallel Generation method with the Pre-Filling strategy (IPGPF). Event roles in an event record are generated in parallel to avoid order selection, and the event records are iteratively generated to utilize historical results. Experiments on two public datasets show our IPGPF improves 11.7 F1 than previous parallel models and up to 5.1 F1 than auto-regressive models under the control variable settings. Moreover, our enhanced IPGPF outperforms other entity-enhanced models and achieves new state-of-the-art performance.

An Iteratively Parallel Generation Method with the Pre-Filling Strategy for Document-level Event Extraction

emnlp-main.668

Poster

https://aclanthology.org/2023.emnlp-main.669.bib

https://aclanthology.org/2023.emnlp-main.669/

@inproceedings{cheng-etal-2023-compost, title = "{C}o{MP}os{T}: Characterizing and Evaluating Caricature in {LLM} Simulations", author = "Cheng, Myra and Piccardi, Tiziano and Yang, Diyi", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.669", doi = "10.18653/v1/2023.emnlp-main.669", pages = "10853--10875", abstract = "Recent work has aimed to capture nuances of human behavior by using LLMs to simulate responses from particular demographics in settings like social science experiments and public opinion surveys. However, there are currently no established ways to discuss or evaluate the quality of such LLM simulations. Moreover, there is growing concern that these LLM simulations are flattened caricatures of the personas that they aim to simulate, failing to capture the multidimensionality of people and perpetuating stereotypes. To bridge these gaps, we present CoMPosT, a framework to characterize LLM simulations using four dimensions: Context, Model, Persona, and Topic. We use this framework to measure open-ended LLM simulations{'} susceptibility to caricature, defined via two criteria: individuation and exaggeration. We evaluate the level of caricature in scenarios from existing work on LLM simulations. We find that for GPT-4, simulations of certain demographics (political and marginalized groups) and topics (general, uncontroversial) are highly susceptible to caricature.", }

Recent work has aimed to capture nuances of human behavior by using LLMs to simulate responses from particular demographics in settings like social science experiments and public opinion surveys. However, there are currently no established ways to discuss or evaluate the quality of such LLM simulations. Moreover, there is growing concern that these LLM simulations are flattened caricatures of the personas that they aim to simulate, failing to capture the multidimensionality of people and perpetuating stereotypes. To bridge these gaps, we present CoMPosT, a framework to characterize LLM simulations using four dimensions: Context, Model, Persona, and Topic. We use this framework to measure open-ended LLM simulations{'} susceptibility to caricature, defined via two criteria: individuation and exaggeration. We evaluate the level of caricature in scenarios from existing work on LLM simulations. We find that for GPT-4, simulations of certain demographics (political and marginalized groups) and topics (general, uncontroversial) are highly susceptible to caricature.

CoMPosT: Characterizing and Evaluating Caricature in LLM Simulations

emnlp-main.669

2310.11501

Poster

https://aclanthology.org/2023.emnlp-main.670.bib

https://aclanthology.org/2023.emnlp-main.670/

@inproceedings{huang-etal-2023-reduce, title = "Reduce Human Labor On Evaluating Conversational Information Retrieval System: A Human-Machine Collaboration Approach", author = "Huang, Chen and Qin, Peixin and Lei, Wenqiang and Lv, Jiancheng", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.670", doi = "10.18653/v1/2023.emnlp-main.670", pages = "10876--10891", abstract = "Evaluating conversational information retrieval (CIR) systems is a challenging task that requires a significant amount of human labor for annotation. It is imperative to invest significant effort into researching more labor-effective methods for evaluating CIR systems. To touch upon this challenge, we take the first step to involve active testing in CIR evaluation and propose a novel method, called HomCoE. It strategically selects a few data for human annotation, then calibrates the evaluation results to eliminate evaluation biases. As such, it makes an accurate evaluation of the CIR system at low human labor. We experimentally reveal that it consumes less than 1{\%} of human labor and achieves a consistency rate of 95{\%}-99{\%} with human evaluation results. This emphasizes the superiority of our method over other baselines.", }

Evaluating conversational information retrieval (CIR) systems is a challenging task that requires a significant amount of human labor for annotation. It is imperative to invest significant effort into researching more labor-effective methods for evaluating CIR systems. To touch upon this challenge, we take the first step to involve active testing in CIR evaluation and propose a novel method, called HomCoE. It strategically selects a few data for human annotation, then calibrates the evaluation results to eliminate evaluation biases. As such, it makes an accurate evaluation of the CIR system at low human labor. We experimentally reveal that it consumes less than 1{\%} of human labor and achieves a consistency rate of 95{\%}-99{\%} with human evaluation results. This emphasizes the superiority of our method over other baselines.

Reduce Human Labor On Evaluating Conversational Information Retrieval System: A Human-Machine Collaboration Approach

emnlp-main.670

Poster

https://aclanthology.org/2023.emnlp-main.671.bib

https://aclanthology.org/2023.emnlp-main.671/

@inproceedings{bruton-beloucif-2023-bertie, title = "{BERT}ie Bott{'}s Every Flavor Labels: A Tasty Introduction to Semantic Role Labeling for {G}alician", author = "Bruton, Micaella and Beloucif, Meriem", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.671", doi = "10.18653/v1/2023.emnlp-main.671", pages = "10892--10902", abstract = "In this paper, we leverage existing corpora, WordNet, and dependency parsing to build the first Galician dataset for training semantic role labeling systems in an effort to expand available NLP resources. Additionally, we introduce verb indexing, a new pre-processing method, which helps increase the performance when semantically parsing highly-complex sentences. We use transfer-learning to test both the resource and the verb indexing method. Our results show that the effects of verb indexing were amplified in scenarios where the model was both pre-trained and fine-tuned on datasets utilizing the method, but improvements are also noticeable when only used during fine-tuning. The best-performing Galician SRL model achieved an f1 score of 0.74, introducing a baseline for future Galician SRL systems. We also tested our method on Spanish where we achieved an f1 score of 0.83, outperforming the baseline set by the 2009 CoNLL Shared Task by 0.025 showing the merits of our verb indexing method for pre-processing.", }

In this paper, we leverage existing corpora, WordNet, and dependency parsing to build the first Galician dataset for training semantic role labeling systems in an effort to expand available NLP resources. Additionally, we introduce verb indexing, a new pre-processing method, which helps increase the performance when semantically parsing highly-complex sentences. We use transfer-learning to test both the resource and the verb indexing method. Our results show that the effects of verb indexing were amplified in scenarios where the model was both pre-trained and fine-tuned on datasets utilizing the method, but improvements are also noticeable when only used during fine-tuning. The best-performing Galician SRL model achieved an f1 score of 0.74, introducing a baseline for future Galician SRL systems. We also tested our method on Spanish where we achieved an f1 score of 0.83, outperforming the baseline set by the 2009 CoNLL Shared Task by 0.025 showing the merits of our verb indexing method for pre-processing.

BERTie Bott's Every Flavor Labels: A Tasty Introduction to Semantic Role Labeling for Galician

emnlp-main.671

Poster

https://aclanthology.org/2023.emnlp-main.672.bib

https://aclanthology.org/2023.emnlp-main.672/

@inproceedings{huang-etal-2023-program, title = "Program Translation via Code Distillation", author = "Huang, Yufan and Qi, Mengnan and Yao, Yongqiang and Wang, Maoquan and Gu, Bin and Clement, Colin and Sundaresan, Neel", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.672", doi = "10.18653/v1/2023.emnlp-main.672", pages = "10903--10914", abstract = "Software version migration and program translation are an important and costly part of the lifecycle of large codebases. Traditional machine translation relies on parallel corpora for supervised translation, which is not feasible for program translation due to a dearth of aligned data. Recent unsupervised neural machine translation techniques have overcome data limitations by included techniques such as back translation and low level compiler intermediate representations (IR). These methods face significant challenges due to the noise in code snippet alignment and the diversity of IRs respectively. In this paper we propose a novel model called Code Distillation (CoDist) whereby we capture the semantic and structural equivalence of code in a language agnostic intermediate representation. Distilled code serves as a translation pivot for any programming language, leading by construction to parallel corpora which scale to all available source code by simply applying the distillation compiler. We demonstrate that our approach achieves state-of-the-art performance on CodeXGLUE and TransCoder GeeksForGeeks translation benchmarks, with an average absolute increase of 12.7{\%} on the TransCoder GeeksforGeeks translation benchmark compare to TransCoder-ST.", }

Software version migration and program translation are an important and costly part of the lifecycle of large codebases. Traditional machine translation relies on parallel corpora for supervised translation, which is not feasible for program translation due to a dearth of aligned data. Recent unsupervised neural machine translation techniques have overcome data limitations by included techniques such as back translation and low level compiler intermediate representations (IR). These methods face significant challenges due to the noise in code snippet alignment and the diversity of IRs respectively. In this paper we propose a novel model called Code Distillation (CoDist) whereby we capture the semantic and structural equivalence of code in a language agnostic intermediate representation. Distilled code serves as a translation pivot for any programming language, leading by construction to parallel corpora which scale to all available source code by simply applying the distillation compiler. We demonstrate that our approach achieves state-of-the-art performance on CodeXGLUE and TransCoder GeeksForGeeks translation benchmarks, with an average absolute increase of 12.7{\%} on the TransCoder GeeksforGeeks translation benchmark compare to TransCoder-ST.

Program Translation via Code Distillation

emnlp-main.672

2310.11476

Poster

https://aclanthology.org/2023.emnlp-main.673.bib

https://aclanthology.org/2023.emnlp-main.673/

@inproceedings{zhang-etal-2023-famesumm, title = "{F}a{M}e{S}umm: Investigating and Improving Faithfulness of Medical Summarization", author = "Zhang, Nan and Zhang, Yusen and Guo, Wu and Mitra, Prasenjit and Zhang, Rui", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.673", doi = "10.18653/v1/2023.emnlp-main.673", pages = "10915--10931", abstract = "Summaries of medical text shall be faithful by being consistent and factual with source inputs, which is an important but understudied topic for safety and efficiency in healthcare. In this paper, we investigate and improve faithfulness in summarization on a broad range of medical summarization tasks. Our investigation reveals that current summarization models often produce unfaithful outputs for medical input text. We then introduce FaMeSumm, a framework to improve faithfulness by fine-tuning pre-trained language models based on medical knowledge. FaMeSumm performs contrastive learning on designed sets of faithful and unfaithful summaries, and it incorporates medical terms and their contexts to encourage faithful generation of medical terms. We conduct comprehensive experiments on three datasets in two languages: health question and radiology report summarization datasets in English, and a patient-doctor dialogue dataset in Chinese. Results demonstrate that FaMeSumm is flexible and effective by delivering consistent improvements over mainstream language models such as BART, T5, mT5, and PEGASUS, yielding state-of-the-art performances on metrics for faithfulness and general quality. Human evaluation by doctors also shows that FaMeSumm generates more faithful outputs. Our code is available at https://github.com/psunlpgroup/FaMeSumm.", }

Summaries of medical text shall be faithful by being consistent and factual with source inputs, which is an important but understudied topic for safety and efficiency in healthcare. In this paper, we investigate and improve faithfulness in summarization on a broad range of medical summarization tasks. Our investigation reveals that current summarization models often produce unfaithful outputs for medical input text. We then introduce FaMeSumm, a framework to improve faithfulness by fine-tuning pre-trained language models based on medical knowledge. FaMeSumm performs contrastive learning on designed sets of faithful and unfaithful summaries, and it incorporates medical terms and their contexts to encourage faithful generation of medical terms. We conduct comprehensive experiments on three datasets in two languages: health question and radiology report summarization datasets in English, and a patient-doctor dialogue dataset in Chinese. Results demonstrate that FaMeSumm is flexible and effective by delivering consistent improvements over mainstream language models such as BART, T5, mT5, and PEGASUS, yielding state-of-the-art performances on metrics for faithfulness and general quality. Human evaluation by doctors also shows that FaMeSumm generates more faithful outputs. Our code is available at https://github.com/psunlpgroup/FaMeSumm.

FaMeSumm: Investigating and Improving Faithfulness of Medical Summarization

emnlp-main.673

2311.02271

Poster

https://aclanthology.org/2023.emnlp-main.674.bib

https://aclanthology.org/2023.emnlp-main.674/

@inproceedings{geng-etal-2023-grammar, title = "Grammar-Constrained Decoding for Structured {NLP} Tasks without Finetuning", author = "Geng, Saibo and Josifoski, Martin and Peyrard, Maxime and West, Robert", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.674", doi = "10.18653/v1/2023.emnlp-main.674", pages = "10932--10952", abstract = "Despite their impressive performance, large language models (LMs) still struggle with reliably generating complex output structures when not finetuned to follow the required output format exactly. To address this issue, grammar-constrained decoding (GCD) can be used to control the generation of LMs, guaranteeing that the output follows a given structure. Most existing GCD methods are, however, limited to specific tasks, such as parsing or code generation. In this work, we demonstrate that formal grammars can describe the output space for a much wider range of tasks and argue that GCD can serve as a unified framework for structured NLP tasks in general. For increased flexibility, we introduce input-dependent grammars, which allow the grammar to depend on the input and thus enable the generation of different output structures for different inputs. We then empirically demonstrate the power and flexibility of GCD-enhanced LMs on (1) information extraction, (2) entity disambiguation, and (3) constituency parsing. Our results indicate that grammar-constrained LMs substantially outperform unconstrained LMs or even beat task-specific finetuned models. Grammar constraints thus hold great promise for harnessing off-the-shelf LMs for a wide range of structured NLP tasks, especially where training data is scarce or finetuning is expensive. Code and data: https://github.com/epfl-dlab/GCD.", }

Despite their impressive performance, large language models (LMs) still struggle with reliably generating complex output structures when not finetuned to follow the required output format exactly. To address this issue, grammar-constrained decoding (GCD) can be used to control the generation of LMs, guaranteeing that the output follows a given structure. Most existing GCD methods are, however, limited to specific tasks, such as parsing or code generation. In this work, we demonstrate that formal grammars can describe the output space for a much wider range of tasks and argue that GCD can serve as a unified framework for structured NLP tasks in general. For increased flexibility, we introduce input-dependent grammars, which allow the grammar to depend on the input and thus enable the generation of different output structures for different inputs. We then empirically demonstrate the power and flexibility of GCD-enhanced LMs on (1) information extraction, (2) entity disambiguation, and (3) constituency parsing. Our results indicate that grammar-constrained LMs substantially outperform unconstrained LMs or even beat task-specific finetuned models. Grammar constraints thus hold great promise for harnessing off-the-shelf LMs for a wide range of structured NLP tasks, especially where training data is scarce or finetuning is expensive. Code and data: https://github.com/epfl-dlab/GCD.

Grammar-Constrained Decoding for Structured NLP Tasks without Finetuning

emnlp-main.674

2305.13971

https://huggingface.co/papers/2305.13971

Poster

https://aclanthology.org/2023.emnlp-main.675.bib

https://aclanthology.org/2023.emnlp-main.675/

@inproceedings{yu-2023-systematic, title = "Systematic word meta-sense extension", author = "Yu, Lei", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.675", doi = "10.18653/v1/2023.emnlp-main.675", pages = "10953--10966", abstract = "The meaning of polysemous words often varies in a highly productive yet predictable way. Generalizing the regularity between conventional senses to derive novel word meaning is crucial for automated processing of non-literal language uses such as figurative expressions. We introduce a novel task called systematic word meta-sense extension (SWORME) to test and improve language models{'} ability to extend word meaning to denote new semantic domains (also called meta-senses) that bear regular semantic relations with existing senses. We found that language models prefer incremental lexical semantic change toward conceptually similar meta-senses such as logical metonymy, and are much worse at predicting highly non-literal meaning extensions such as metaphors. We propose a novel analogy-based method of word meaning extension, and show that it effectively improves language model systematicity in making both gradual and radical types of meta-sense extension. We further demonstrate that learning systematic meta-sense extensions benefits language models on multiple benchmarks of figurative language understanding.", }

The meaning of polysemous words often varies in a highly productive yet predictable way. Generalizing the regularity between conventional senses to derive novel word meaning is crucial for automated processing of non-literal language uses such as figurative expressions. We introduce a novel task called systematic word meta-sense extension (SWORME) to test and improve language models{'} ability to extend word meaning to denote new semantic domains (also called meta-senses) that bear regular semantic relations with existing senses. We found that language models prefer incremental lexical semantic change toward conceptually similar meta-senses such as logical metonymy, and are much worse at predicting highly non-literal meaning extensions such as metaphors. We propose a novel analogy-based method of word meaning extension, and show that it effectively improves language model systematicity in making both gradual and radical types of meta-sense extension. We further demonstrate that learning systematic meta-sense extensions benefits language models on multiple benchmarks of figurative language understanding.

Systematic word meta-sense extension

emnlp-main.675

2311.13029

Oral

https://aclanthology.org/2023.emnlp-main.676.bib

https://aclanthology.org/2023.emnlp-main.676/

@inproceedings{xiao-etal-2023-evaluating-evaluation, title = "Evaluating Evaluation Metrics: A Framework for Analyzing {NLG} Evaluation Metrics using Measurement Theory", author = "Xiao, Ziang and Zhang, Susu and Lai, Vivian and Liao, Q. Vera", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.676", doi = "10.18653/v1/2023.emnlp-main.676", pages = "10967--10982", abstract = "We address a fundamental challenge in Natural Language Generation (NLG) model evaluation{---}the design and evaluation of evaluation metrics. Recognizing the limitations of existing automatic metrics and noises from how current human evaluation was conducted, we propose MetricEval, a framework informed by measurement theory, the foundation of educational test design, for conceptualizing and evaluating the reliability and validity of NLG evaluation metrics. The framework formalizes the source of measurement error and offers statistical tools for evaluating evaluation metrics based on empirical data. With our framework, one can quantify the uncertainty of the metrics to better interpret the result. To exemplify the use of our framework in practice, we analyzed a set of evaluation metrics for summarization and identified issues related to conflated validity structure in human-eval and reliability in LLM-based metrics. Through MetricEval, we aim to promote the design, evaluation, and interpretation of valid and reliable metrics to advance robust and effective NLG models.", }

We address a fundamental challenge in Natural Language Generation (NLG) model evaluation{---}the design and evaluation of evaluation metrics. Recognizing the limitations of existing automatic metrics and noises from how current human evaluation was conducted, we propose MetricEval, a framework informed by measurement theory, the foundation of educational test design, for conceptualizing and evaluating the reliability and validity of NLG evaluation metrics. The framework formalizes the source of measurement error and offers statistical tools for evaluating evaluation metrics based on empirical data. With our framework, one can quantify the uncertainty of the metrics to better interpret the result. To exemplify the use of our framework in practice, we analyzed a set of evaluation metrics for summarization and identified issues related to conflated validity structure in human-eval and reliability in LLM-based metrics. Through MetricEval, we aim to promote the design, evaluation, and interpretation of valid and reliable metrics to advance robust and effective NLG models.

Evaluating Evaluation Metrics: A Framework for Analyzing NLG Evaluation Metrics using Measurement Theory

emnlp-main.676

2305.14889

Poster

https://aclanthology.org/2023.emnlp-main.677.bib

https://aclanthology.org/2023.emnlp-main.677/

@inproceedings{fu-etal-2023-revisiting, title = "Revisiting the Knowledge Injection Frameworks", author = "Fu, Peng and Zhang, Yiming and Wang, Haobo and Qiu, Weikang and Zhao, Junbo", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.677", doi = "10.18653/v1/2023.emnlp-main.677", pages = "10983--10997", abstract = "In recent years, large language models (LLMs), such as GPTs, have attained great impact worldwide. However, how to adapt these LLMs to better suit the vertical domain-specific tasks by utilizing external knowledge remains not completely solved. Indeed, there have emerged a few works on this line where most of them rely on an alignment heuristic that is built to inject the corresponding knowledge tuple into the associated text sample. However, despite the promise, we identify a pivotal problem in this work ubiquitously. Simply put, we find that injecting unaligned (i.e., random) knowledge tuple into the LLMs achieves comparable (and sometimes better) results than the aligned knowledge being injected. We therefore take a thorough investigation of this frustrating finding on a variety of related prior work and further provide a chain of potential interpretations for the phenomenon. Based on all that, we offer a simple remediated technique. Briefly, the core of this technique roots in an ideological emphasis on the pruning and purification of the external knowledge base to be injected into LLMs. At last, we show that by integrating this technique into most (if not all) knowledge injection frameworks and recent LLMs, it manages to overcome the aforementioned sanity problem and further pushes the boundary of the performance of the domain-adaptive LLMs.", }

In recent years, large language models (LLMs), such as GPTs, have attained great impact worldwide. However, how to adapt these LLMs to better suit the vertical domain-specific tasks by utilizing external knowledge remains not completely solved. Indeed, there have emerged a few works on this line where most of them rely on an alignment heuristic that is built to inject the corresponding knowledge tuple into the associated text sample. However, despite the promise, we identify a pivotal problem in this work ubiquitously. Simply put, we find that injecting unaligned (i.e., random) knowledge tuple into the LLMs achieves comparable (and sometimes better) results than the aligned knowledge being injected. We therefore take a thorough investigation of this frustrating finding on a variety of related prior work and further provide a chain of potential interpretations for the phenomenon. Based on all that, we offer a simple remediated technique. Briefly, the core of this technique roots in an ideological emphasis on the pruning and purification of the external knowledge base to be injected into LLMs. At last, we show that by integrating this technique into most (if not all) knowledge injection frameworks and recent LLMs, it manages to overcome the aforementioned sanity problem and further pushes the boundary of the performance of the domain-adaptive LLMs.

Revisiting the Knowledge Injection Frameworks

emnlp-main.677

2311.01150

Poster

https://aclanthology.org/2023.emnlp-main.678.bib

https://aclanthology.org/2023.emnlp-main.678/

@inproceedings{kane-schubert-2023-repeatedly, title = "We Are What We Repeatedly Do: Inducing and Deploying Habitual Schemas in Persona-Based Responses", author = "Kane, Benjamin and Schubert, Lenhart", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.678", doi = "10.18653/v1/2023.emnlp-main.678", pages = "10998--11016", abstract = "Many practical applications of dialogue technology require the generation of responses according to a particular developer-specified persona. While a variety of personas can be elicited from recent large language models, the opaqueness and unpredictability of these models make it desirable to be able to specify personas in an explicit form. In previous work, personas have typically been represented as sets of one-off pieces of self-knowledge that are retrieved by the dialogue system for use in generation. However, in realistic human conversations, personas are often revealed through story-like narratives that involve rich habitual knowledge {--} knowledge about kinds of events that an agent often participates in (e.g., work activities, hobbies, sporting activities, favorite entertainments, etc.), including typical goals, sub-events, preconditions, and postconditions of those events. We capture such habitual knowledge using an explicit schema representation, and propose an approach to dialogue generation that retrieves relevant schemas to condition a large language model to generate persona-based responses. Furthermore, we demonstrate a method for bootstrapping the creation of such schemas by first generating generic passages from a set of simple facts, and then inducing schemas from the generated passages.", }

Many practical applications of dialogue technology require the generation of responses according to a particular developer-specified persona. While a variety of personas can be elicited from recent large language models, the opaqueness and unpredictability of these models make it desirable to be able to specify personas in an explicit form. In previous work, personas have typically been represented as sets of one-off pieces of self-knowledge that are retrieved by the dialogue system for use in generation. However, in realistic human conversations, personas are often revealed through story-like narratives that involve rich habitual knowledge {--} knowledge about kinds of events that an agent often participates in (e.g., work activities, hobbies, sporting activities, favorite entertainments, etc.), including typical goals, sub-events, preconditions, and postconditions of those events. We capture such habitual knowledge using an explicit schema representation, and propose an approach to dialogue generation that retrieves relevant schemas to condition a large language model to generate persona-based responses. Furthermore, we demonstrate a method for bootstrapping the creation of such schemas by first generating generic passages from a set of simple facts, and then inducing schemas from the generated passages.

We Are What We Repeatedly Do: Inducing and Deploying Habitual Schemas in Persona-Based Responses

emnlp-main.678

2310.06245

https://huggingface.co/papers/2310.06245

Poster

https://aclanthology.org/2023.emnlp-main.679.bib

https://aclanthology.org/2023.emnlp-main.679/

@inproceedings{jia-etal-2023-zero, title = "Zero-shot Faithfulness Evaluation for Text Summarization with Foundation Language Model", author = "Jia, Qi and Ren, Siyu and Liu, Yizhu and Zhu, Kenny", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.679", doi = "10.18653/v1/2023.emnlp-main.679", pages = "11017--11031", abstract = "Despite tremendous improvements in natural language generation, summarization models still suffer from the unfaithfulness issue. Previous work evaluates faithfulness either using models trained on the other tasks or in-domain synthetic data, or prompting a large model such as ChatGPT. This paper proposes to do zero-shot faithfulness evaluation simply with a moderately-sized foundation language model. We introduce a new metric FFLM, which is a combination of probability changes based on the intuition that prefixing a piece of text that is consistent with the output will increase the probability of predicting the output. Experiments show that FFLM performs competitively with or even outperforms ChatGPT on both inconsistency detection and faithfulness rating with 24x fewer parameters. FFLM also achieves improvements over other strong baselines.", }

Despite tremendous improvements in natural language generation, summarization models still suffer from the unfaithfulness issue. Previous work evaluates faithfulness either using models trained on the other tasks or in-domain synthetic data, or prompting a large model such as ChatGPT. This paper proposes to do zero-shot faithfulness evaluation simply with a moderately-sized foundation language model. We introduce a new metric FFLM, which is a combination of probability changes based on the intuition that prefixing a piece of text that is consistent with the output will increase the probability of predicting the output. Experiments show that FFLM performs competitively with or even outperforms ChatGPT on both inconsistency detection and faithfulness rating with 24x fewer parameters. FFLM also achieves improvements over other strong baselines.

Zero-shot Faithfulness Evaluation for Text Summarization with Foundation Language Model

emnlp-main.679

2310.11648

Oral

https://aclanthology.org/2023.emnlp-main.680.bib

https://aclanthology.org/2023.emnlp-main.680/

@inproceedings{kim-etal-2023-taskweb, title = "{T}ask{W}eb: Selecting Better Source Tasks for Multi-task {NLP}", author = "Kim, Joongwon and Asai, Akari and Ilharco, Gabriel and Hajishirzi, Hannaneh", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.680", doi = "10.18653/v1/2023.emnlp-main.680", pages = "11032--11052", abstract = "Recent work in NLP has shown promising results in training models on large amounts of tasks to achieve better generalization. However, it is not well-understood how tasks are related, and how helpful training tasks can be chosen for a new task. In this work, we investigate whether knowing task relationships via pairwise task transfer improves choosing one or more source tasks that help to learn a new target task. We provide TaskWeb, a large-scale benchmark of pairwise task transfers for 22 NLP tasks using three different model types, sizes, and adaptation methods, spanning about 25,000 experiments. Then, we design a new method TaskShop based on our analysis of TaskWeb. TaskShop uses TaskWeb to estimate the benefit of using a source task for learning a new target task, and to choose a subset of helpful training tasks for multi-task training. Our method improves overall rankings and top-k precision of source tasks by 10{\%} and 38{\%}, respectively. We also use TaskShop to build much smaller multi-task training sets that improve zero-shot performances across 11 different target tasks by at least 4.3{\%}.", }

Recent work in NLP has shown promising results in training models on large amounts of tasks to achieve better generalization. However, it is not well-understood how tasks are related, and how helpful training tasks can be chosen for a new task. In this work, we investigate whether knowing task relationships via pairwise task transfer improves choosing one or more source tasks that help to learn a new target task. We provide TaskWeb, a large-scale benchmark of pairwise task transfers for 22 NLP tasks using three different model types, sizes, and adaptation methods, spanning about 25,000 experiments. Then, we design a new method TaskShop based on our analysis of TaskWeb. TaskShop uses TaskWeb to estimate the benefit of using a source task for learning a new target task, and to choose a subset of helpful training tasks for multi-task training. Our method improves overall rankings and top-k precision of source tasks by 10{\%} and 38{\%}, respectively. We also use TaskShop to build much smaller multi-task training sets that improve zero-shot performances across 11 different target tasks by at least 4.3{\%}.

TaskWeb: Selecting Better Source Tasks for Multi-task NLP

emnlp-main.680

Poster

https://aclanthology.org/2023.emnlp-main.681.bib

https://aclanthology.org/2023.emnlp-main.681/

@inproceedings{jeon-etal-2023-improving, title = "Improving Bias Mitigation through Bias Experts in Natural Language Understanding", author = "Jeon, Eojin and Lee, Mingyu and Park, Juhyeong and Kim, Yeachan and Mok, Wing-Lam and Lee, SangKeun", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.681", doi = "10.18653/v1/2023.emnlp-main.681", pages = "11053--11066", abstract = "Biases in the dataset often enable the model to achieve high performance on in-distribution data, while poorly performing on out-of-distribution data. To mitigate the detrimental effect of the bias on the networks, previous works have proposed debiasing methods that down-weight the biased examples identified by an auxiliary model, which is trained with explicit bias labels. However, finding a type of bias in datasets is a costly process. Therefore, recent studies have attempted to make the auxiliary model biased without the guidance (or annotation) of bias labels, by constraining the model{'}s training environment or the capability of the model itself. Despite the promising debiasing results of recent works, the multi-class learning objective, which has been naively used to train the auxiliary model, may harm the bias mitigation effect due to its regularization effect and competitive nature across classes. As an alternative, we propose a new debiasing framework that introduces binary classifiers between the auxiliary model and the main model, coined bias experts. Specifically, each bias expert is trained on a binary classification task derived from the multi-class classification task via the One-vs-Rest approach. Experimental results demonstrate that our proposed strategy improves the bias identification ability of the auxiliary model. Consequently, our debiased model consistently outperforms the state-of-the-art on various challenge datasets.", }

Biases in the dataset often enable the model to achieve high performance on in-distribution data, while poorly performing on out-of-distribution data. To mitigate the detrimental effect of the bias on the networks, previous works have proposed debiasing methods that down-weight the biased examples identified by an auxiliary model, which is trained with explicit bias labels. However, finding a type of bias in datasets is a costly process. Therefore, recent studies have attempted to make the auxiliary model biased without the guidance (or annotation) of bias labels, by constraining the model{'}s training environment or the capability of the model itself. Despite the promising debiasing results of recent works, the multi-class learning objective, which has been naively used to train the auxiliary model, may harm the bias mitigation effect due to its regularization effect and competitive nature across classes. As an alternative, we propose a new debiasing framework that introduces binary classifiers between the auxiliary model and the main model, coined bias experts. Specifically, each bias expert is trained on a binary classification task derived from the multi-class classification task via the One-vs-Rest approach. Experimental results demonstrate that our proposed strategy improves the bias identification ability of the auxiliary model. Consequently, our debiased model consistently outperforms the state-of-the-art on various challenge datasets.

Improving Bias Mitigation through Bias Experts in Natural Language Understanding

emnlp-main.681

2312.03577

Oral

https://aclanthology.org/2023.emnlp-main.682.bib

https://aclanthology.org/2023.emnlp-main.682/

@inproceedings{goel-etal-2023-semi, title = "Semi-supervised multimodal coreference resolution in image narrations", author = "Goel, Arushi and Fernando, Basura and Keller, Frank and Bilen, Hakan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.682", doi = "10.18653/v1/2023.emnlp-main.682", pages = "11067--11081", abstract = "In this paper, we study multimodal coreference resolution, specifically where a longer descriptive text, i.e., a narration is paired with an image. This poses significant challenges due to fine-grained image-text alignment, inherent ambiguity present in narrative language, and unavailability of large annotated training sets. To tackle these challenges, we present a data efficient semi-supervised approach that utilizes image-narration pairs to resolve coreferences and narrative grounding in a multimodal context. Our approach incorporates losses for both labeled and unlabeled data within a cross-modal framework. Our evaluation shows that the proposed approach outperforms strong baselines both quantitatively and qualitatively, for the tasks of coreference resolution and narrative grounding.", }

In this paper, we study multimodal coreference resolution, specifically where a longer descriptive text, i.e., a narration is paired with an image. This poses significant challenges due to fine-grained image-text alignment, inherent ambiguity present in narrative language, and unavailability of large annotated training sets. To tackle these challenges, we present a data efficient semi-supervised approach that utilizes image-narration pairs to resolve coreferences and narrative grounding in a multimodal context. Our approach incorporates losses for both labeled and unlabeled data within a cross-modal framework. Our evaluation shows that the proposed approach outperforms strong baselines both quantitatively and qualitatively, for the tasks of coreference resolution and narrative grounding.

Semi-supervised multimodal coreference resolution in image narrations

emnlp-main.682

2310.13619

Poster

https://aclanthology.org/2023.emnlp-main.683.bib

https://aclanthology.org/2023.emnlp-main.683/

@inproceedings{zhou-etal-2023-predictive, title = "A Predictive Factor Analysis of Social Biases and Task-Performance in Pretrained Masked Language Models", author = "Zhou, Yi and Camacho-Collados, Jose and Bollegala, Danushka", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.683", doi = "10.18653/v1/2023.emnlp-main.683", pages = "11082--11100", abstract = "Various types of social biases have been reported with pretrained Masked Language Models (MLMs) in prior work. However, multiple underlying factors are associated with an MLM such as its model size, size of the training data, training objectives, the domain from which pretraining data is sampled, tokenization, and languages present in the pretrained corpora, to name a few. It remains unclear as to which of those factors influence social biases that are learned by MLMs. To study the relationship between model factors and the social biases learned by an MLM, as well as the downstream task performance of the model, we conduct a comprehensive study over 39 pretrained MLMs covering different model sizes, training objectives, tokenization methods, training data domains and languages. Our results shed light on important factors often neglected in prior literature, such as tokenization or model objectives.", }

Various types of social biases have been reported with pretrained Masked Language Models (MLMs) in prior work. However, multiple underlying factors are associated with an MLM such as its model size, size of the training data, training objectives, the domain from which pretraining data is sampled, tokenization, and languages present in the pretrained corpora, to name a few. It remains unclear as to which of those factors influence social biases that are learned by MLMs. To study the relationship between model factors and the social biases learned by an MLM, as well as the downstream task performance of the model, we conduct a comprehensive study over 39 pretrained MLMs covering different model sizes, training objectives, tokenization methods, training data domains and languages. Our results shed light on important factors often neglected in prior literature, such as tokenization or model objectives.

A Predictive Factor Analysis of Social Biases and Task-Performance in Pretrained Masked Language Models

emnlp-main.683

2310.12936

Poster

https://aclanthology.org/2023.emnlp-main.684.bib

https://aclanthology.org/2023.emnlp-main.684/

@inproceedings{goffredo-etal-2023-argument, title = "Argument-based Detection and Classification of Fallacies in Political Debates", author = "Goffredo, Pierpaolo and Chaves, Mariana and Villata, Serena and Cabrio, Elena", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.684", doi = "10.18653/v1/2023.emnlp-main.684", pages = "11101--11112", abstract = "Fallacies are arguments that employ faulty reasoning. Given their persuasive and seemingly valid nature, fallacious arguments are often used in political debates. Employing these misleading arguments in politics can have detrimental consequences for society, since they can lead to inaccurate conclusions and invalid inferences from the public opinion and the policymakers. Automatically detecting and classifying fallacious arguments represents therefore a crucial challenge to limit the spread of misleading or manipulative claims and promote a more informed and healthier political discourse. Our contribution to address this challenging task is twofold. First, we extend the ElecDeb60To16 dataset of U.S. presidential debates annotated with fallacious arguments, by incorporating the most recent Trump-Biden presidential debate. We include updated token-level annotations, incorporating argumentative components (i.e., claims and premises), the relations between these components (i.e., support and attack), and six categories of fallacious arguments (i.e., Ad Hominem, Appeal to Authority, Appeal to Emotion, False Cause, Slippery Slope, and Slogans). Second, we perform the twofold task of fallacious argument detection and classification by defining neural network architectures based on Transformers models, combining text, argumentative features, and engineered features. Our results show the advantages of complementing transformer-generated text representations with non-text features.", }

Fallacies are arguments that employ faulty reasoning. Given their persuasive and seemingly valid nature, fallacious arguments are often used in political debates. Employing these misleading arguments in politics can have detrimental consequences for society, since they can lead to inaccurate conclusions and invalid inferences from the public opinion and the policymakers. Automatically detecting and classifying fallacious arguments represents therefore a crucial challenge to limit the spread of misleading or manipulative claims and promote a more informed and healthier political discourse. Our contribution to address this challenging task is twofold. First, we extend the ElecDeb60To16 dataset of U.S. presidential debates annotated with fallacious arguments, by incorporating the most recent Trump-Biden presidential debate. We include updated token-level annotations, incorporating argumentative components (i.e., claims and premises), the relations between these components (i.e., support and attack), and six categories of fallacious arguments (i.e., Ad Hominem, Appeal to Authority, Appeal to Emotion, False Cause, Slippery Slope, and Slogans). Second, we perform the twofold task of fallacious argument detection and classification by defining neural network architectures based on Transformers models, combining text, argumentative features, and engineered features. Our results show the advantages of complementing transformer-generated text representations with non-text features.

Argument-based Detection and Classification of Fallacies in Political Debates

emnlp-main.684

Oral

https://aclanthology.org/2023.emnlp-main.685.bib

https://aclanthology.org/2023.emnlp-main.685/

@inproceedings{zhu-etal-2023-collaborative, title = "Collaborative Generative {AI}: Integrating {GPT}-k for Efficient Editing in Text-to-Image Generation", author = "Zhu, Wanrong and Wang, Xinyi and Lu, Yujie and Fu, Tsu-Jui and Wang, Xin and Eckstein, Miguel and Wang, William", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.685", doi = "10.18653/v1/2023.emnlp-main.685", pages = "11113--11122", abstract = "The field of text-to-image (T2I) generation has garnered significant attention both within the research community and among everyday users. Despite the advancements of T2I models, a common issue encountered by users is the need for repetitive editing of input prompts in order to receive a satisfactory image, which is time-consuming and labor-intensive. Given the demonstrated text generation power of large-scale language models, such as GPT-k, we investigate the potential of utilizing such models to improve the prompt editing process for T2I generation. We conduct a series of experiments to compare the common edits made by humans and GPT-k, evaluate the performance of GPT-k in prompting T2I, and examine factors that may influence this process. We found that GPT-k models focus more on inserting modifiers while humans tend to replace words and phrases, which includes changes to the subject matter. Experimental results show that GPT-k are more effective in adjusting modifiers rather than predicting spontaneous changes in the primary subject matters. Adopting the edit suggested by GPT-k models may reduce the percentage of remaining edits by 20-30{\%}.", }

The field of text-to-image (T2I) generation has garnered significant attention both within the research community and among everyday users. Despite the advancements of T2I models, a common issue encountered by users is the need for repetitive editing of input prompts in order to receive a satisfactory image, which is time-consuming and labor-intensive. Given the demonstrated text generation power of large-scale language models, such as GPT-k, we investigate the potential of utilizing such models to improve the prompt editing process for T2I generation. We conduct a series of experiments to compare the common edits made by humans and GPT-k, evaluate the performance of GPT-k in prompting T2I, and examine factors that may influence this process. We found that GPT-k models focus more on inserting modifiers while humans tend to replace words and phrases, which includes changes to the subject matter. Experimental results show that GPT-k are more effective in adjusting modifiers rather than predicting spontaneous changes in the primary subject matters. Adopting the edit suggested by GPT-k models may reduce the percentage of remaining edits by 20-30{\%}.

Collaborative Generative AI: Integrating GPT-k for Efficient Editing in Text-to-Image Generation

emnlp-main.685

2305.11317

Poster

https://aclanthology.org/2023.emnlp-main.686.bib

https://aclanthology.org/2023.emnlp-main.686/

@inproceedings{shavarani-sarkar-2023-spel, title = "{S}p{EL}: Structured Prediction for Entity Linking", author = "Shavarani, Hassan and Sarkar, Anoop", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.686", doi = "10.18653/v1/2023.emnlp-main.686", pages = "11123--11137", abstract = "Entity linking is a prominent thread of research focused on structured data creation by linking spans of text to an ontology or knowledge source. We revisit the use of structured prediction for entity linking which classifies each individual input token as an entity, and aggregates the token predictions. Our system, called SpEL (Structured prediction for Entity Linking) is a state-of-the-art entity linking system that uses some new ideas to apply structured prediction to the task of entity linking including: two refined fine-tuning steps; a context sensitive prediction aggregation strategy; reduction of the size of the model{'}s output vocabulary, and; we address a common problem in entity-linking systems where there is a training vs. inference tokenization mismatch. Our experiments show that we can outperform the state-of-the-art on the commonly used AIDA benchmark dataset for entity linking to Wikipedia. Our method is also very compute efficient in terms of number of parameters and speed of inference.", }

Entity linking is a prominent thread of research focused on structured data creation by linking spans of text to an ontology or knowledge source. We revisit the use of structured prediction for entity linking which classifies each individual input token as an entity, and aggregates the token predictions. Our system, called SpEL (Structured prediction for Entity Linking) is a state-of-the-art entity linking system that uses some new ideas to apply structured prediction to the task of entity linking including: two refined fine-tuning steps; a context sensitive prediction aggregation strategy; reduction of the size of the model{'}s output vocabulary, and; we address a common problem in entity-linking systems where there is a training vs. inference tokenization mismatch. Our experiments show that we can outperform the state-of-the-art on the commonly used AIDA benchmark dataset for entity linking to Wikipedia. Our method is also very compute efficient in terms of number of parameters and speed of inference.

SpEL: Structured Prediction for Entity Linking

emnlp-main.686

Oral

https://aclanthology.org/2023.emnlp-main.687.bib

https://aclanthology.org/2023.emnlp-main.687/

@inproceedings{heinisch-etal-2023-architectural, title = "Architectural Sweet Spots for Modeling Human Label Variation by the Example of Argument Quality: It{'}s Best to Relate Perspectives!", author = "Heinisch, Philipp and Orlikowski, Matthias and Romberg, Julia and Cimiano, Philipp", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.687", doi = "10.18653/v1/2023.emnlp-main.687", pages = "11138--11154", abstract = "Many annotation tasks in natural language processing are highly subjective in that there can be different valid and justified perspectives on what is a proper label for a given example. This also applies to the judgment of argument quality, where the assignment of a single ground truth is often questionable. At the same time, there are generally accepted concepts behind argumentation that form a common ground. To best represent the interplay of individual and shared perspectives, we consider a continuum of approaches ranging from models that fully aggregate perspectives into a majority label to {``}share nothing{''}-architectures in which each annotator is considered in isolation from all other annotators. In between these extremes, inspired by models used in the field of recommender systems, we investigate the extent to which architectures that predict labels for single annotators but include layers that model the relations between different annotators are beneficial. By means of two tasks of argument quality classification (argument concreteness and validity/novelty of conclusions), we show that recommender architectures increase the averaged annotator-individual F1-scores up to 43{\%} over a majority-label model. Our findings indicate that approaches to subjectivity can benefit from relating individual perspectives.", }

Many annotation tasks in natural language processing are highly subjective in that there can be different valid and justified perspectives on what is a proper label for a given example. This also applies to the judgment of argument quality, where the assignment of a single ground truth is often questionable. At the same time, there are generally accepted concepts behind argumentation that form a common ground. To best represent the interplay of individual and shared perspectives, we consider a continuum of approaches ranging from models that fully aggregate perspectives into a majority label to {``}share nothing{''}-architectures in which each annotator is considered in isolation from all other annotators. In between these extremes, inspired by models used in the field of recommender systems, we investigate the extent to which architectures that predict labels for single annotators but include layers that model the relations between different annotators are beneficial. By means of two tasks of argument quality classification (argument concreteness and validity/novelty of conclusions), we show that recommender architectures increase the averaged annotator-individual F1-scores up to 43{\%} over a majority-label model. Our findings indicate that approaches to subjectivity can benefit from relating individual perspectives.

Architectural Sweet Spots for Modeling Human Label Variation by the Example of Argument Quality: It's Best to Relate Perspectives!

emnlp-main.687

Poster

https://aclanthology.org/2023.emnlp-main.688.bib

https://aclanthology.org/2023.emnlp-main.688/

@inproceedings{zhao-etal-2023-explicit, title = "Explicit Planning Helps Language Models in Logical Reasoning", author = "Zhao, Hongyu and Wang, Kangrui and Yu, Mo and Mei, Hongyuan", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.688", doi = "10.18653/v1/2023.emnlp-main.688", pages = "11155--11173", abstract = "Language models have been shown to perform remarkably well on a wide range of natural language processing tasks. In this paper, we propose LEAP, a novel system that uses language models to perform multi-step logical reasoning and incorporates explicit planning into the inference procedure. Explicit planning enables the system to make more informed reasoning decisions at each step by looking ahead into their future effects. Moreover, we propose a training strategy that safeguards the planning process from being led astray by spurious features. Our full system significantly outperforms other competing methods on multiple standard datasets. When using small T5 models as its core selection and deduction components, our system performs competitively compared to GPT-3 despite having only about 1B parameters (i.e., 175 times smaller than GPT-3). When using GPT-3.5, it significantly outperforms chain-of-thought prompting on the challenging PrOntoQA dataset. We have conducted extensive empirical studies to demonstrate that explicit planning plays a crucial role in the system{'}s performance.", }

Language models have been shown to perform remarkably well on a wide range of natural language processing tasks. In this paper, we propose LEAP, a novel system that uses language models to perform multi-step logical reasoning and incorporates explicit planning into the inference procedure. Explicit planning enables the system to make more informed reasoning decisions at each step by looking ahead into their future effects. Moreover, we propose a training strategy that safeguards the planning process from being led astray by spurious features. Our full system significantly outperforms other competing methods on multiple standard datasets. When using small T5 models as its core selection and deduction components, our system performs competitively compared to GPT-3 despite having only about 1B parameters (i.e., 175 times smaller than GPT-3). When using GPT-3.5, it significantly outperforms chain-of-thought prompting on the challenging PrOntoQA dataset. We have conducted extensive empirical studies to demonstrate that explicit planning plays a crucial role in the system{'}s performance.

Explicit Planning Helps Language Models in Logical Reasoning

emnlp-main.688

2303.15714

Oral

https://aclanthology.org/2023.emnlp-main.689.bib

https://aclanthology.org/2023.emnlp-main.689/

@inproceedings{chalamalasetti-etal-2023-clembench, title = "clembench: Using Game Play to Evaluate Chat-Optimized Language Models as Conversational Agents", author = {Chalamalasetti, Kranti and G{\"o}tze, Jana and Hakimov, Sherzod and Madureira, Brielen and Sadler, Philipp and Schlangen, David}, editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.689", doi = "10.18653/v1/2023.emnlp-main.689", pages = "11174--11219", abstract = "Recent work has proposed a methodology for the systematic evaluation of {``}Situated Language Understanding Agents{''} {---} agents that operate in rich linguistic and non-linguistic contexts {---} through testing them in carefully constructed interactive settings. Other recent work has argued that Large Language Models (LLMs), if suitably set up, can be understood as (simulators of) such agents. A connection suggests itself, which this paper explores: Can LLMs be evaluated meaningfully by exposing them to constrained game-like settings that are built to challenge specific capabilities? As a proof of concept, this paper investigates five interaction settings, showing that current chat-optimised LLMs are, to an extent, capable of following game-play instructions. Both this capability and the quality of the game play, measured by how well the objectives of the different games are met, follows the development cycle, with newer models generally performing better. The metrics even for the comparatively simple example games are far from being saturated, suggesting that the proposed instrument will remain to have diagnostic value.", }

Recent work has proposed a methodology for the systematic evaluation of {``}Situated Language Understanding Agents{''} {---} agents that operate in rich linguistic and non-linguistic contexts {---} through testing them in carefully constructed interactive settings. Other recent work has argued that Large Language Models (LLMs), if suitably set up, can be understood as (simulators of) such agents. A connection suggests itself, which this paper explores: Can LLMs be evaluated meaningfully by exposing them to constrained game-like settings that are built to challenge specific capabilities? As a proof of concept, this paper investigates five interaction settings, showing that current chat-optimised LLMs are, to an extent, capable of following game-play instructions. Both this capability and the quality of the game play, measured by how well the objectives of the different games are met, follows the development cycle, with newer models generally performing better. The metrics even for the comparatively simple example games are far from being saturated, suggesting that the proposed instrument will remain to have diagnostic value.

clembench: Using Game Play to Evaluate Chat-Optimized Language Models as Conversational Agents

emnlp-main.689

Poster

https://aclanthology.org/2023.emnlp-main.690.bib

https://aclanthology.org/2023.emnlp-main.690/

@inproceedings{briakou-etal-2023-explaining, title = "Explaining with Contrastive Phrasal Highlighting: A Case Study in Assisting Humans to Detect Translation Differences", author = "Briakou, Eleftheria and Goyal, Navita and Carpuat, Marine", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.690", doi = "10.18653/v1/2023.emnlp-main.690", pages = "11220--11237", abstract = "Explainable NLP techniques primarily explain by answering {``}Which tokens in the input are responsible for this prediction?{''}. We argue that for NLP models that make predictions by comparing two input texts, it is more useful to explain by answering {``}What differences between the two inputs explain this prediction?{''}. We introduce a technique to generate contrastive phrasal highlights that explain the predictions of a semantic divergence model via phrase alignment guided erasure. We show that the resulting highlights match human rationales of cross-lingual semantic differences better than popular post-hoc saliency techniques and that they successfully help people detect fine-grained meaning differences in human translations and critical machine translation errors.", }

Explainable NLP techniques primarily explain by answering {``}Which tokens in the input are responsible for this prediction?{''}. We argue that for NLP models that make predictions by comparing two input texts, it is more useful to explain by answering {``}What differences between the two inputs explain this prediction?{''}. We introduce a technique to generate contrastive phrasal highlights that explain the predictions of a semantic divergence model via phrase alignment guided erasure. We show that the resulting highlights match human rationales of cross-lingual semantic differences better than popular post-hoc saliency techniques and that they successfully help people detect fine-grained meaning differences in human translations and critical machine translation errors.

Explaining with Contrastive Phrasal Highlighting: A Case Study in Assisting Humans to Detect Translation Differences

emnlp-main.690

2312.01582

Poster

https://aclanthology.org/2023.emnlp-main.691.bib

https://aclanthology.org/2023.emnlp-main.691/

@inproceedings{schott-etal-2023-polyglot, title = "Polyglot or Not? Measuring Multilingual Encyclopedic Knowledge in Foundation Models", author = "Schott, Tim and Furman, Daniel and Bhat, Shreshta", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.691", doi = "10.18653/v1/2023.emnlp-main.691", pages = "11238--11253", abstract = "In this work, we assess the ability of foundation models to recall encyclopedic knowledge across a wide range of linguistic contexts. To support this, we: 1) produce a 20-language dataset that contains 303k factual associations paired with counterfactuals, 2) evaluate 5 models in a multilingual test, and 3) benchmark a diverse set of 24 models in an English-only test. Meta{'}s LLaMA achieves the highest scores in both multilingual and English-only evaluations. Yet, an analysis of LLaMA{'}s errors reveals significant limitations in its ability to recall facts in languages other than English, plus difficulties related to the location and gender of fact subjects. Overall, our findings suggest that today{'}s foundation models are far from polyglots.", }

In this work, we assess the ability of foundation models to recall encyclopedic knowledge across a wide range of linguistic contexts. To support this, we: 1) produce a 20-language dataset that contains 303k factual associations paired with counterfactuals, 2) evaluate 5 models in a multilingual test, and 3) benchmark a diverse set of 24 models in an English-only test. Meta{'}s LLaMA achieves the highest scores in both multilingual and English-only evaluations. Yet, an analysis of LLaMA{'}s errors reveals significant limitations in its ability to recall facts in languages other than English, plus difficulties related to the location and gender of fact subjects. Overall, our findings suggest that today{'}s foundation models are far from polyglots.

Polyglot or Not? Measuring Multilingual Encyclopedic Knowledge in Foundation Models

emnlp-main.691

2305.13675

https://huggingface.co/papers/2305.13675

Poster

https://aclanthology.org/2023.emnlp-main.692.bib

https://aclanthology.org/2023.emnlp-main.692/

@inproceedings{pauli-etal-2023-anchoring, title = "Anchoring Fine-tuning of Sentence Transformer with Semantic Label Information for Efficient Truly Few-shot Classification", author = "Pauli, Amalie and Derczynski, Leon and Assent, Ira", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.692", doi = "10.18653/v1/2023.emnlp-main.692", pages = "11254--11264", abstract = "Few-shot classification is a powerful technique, but training requires substantial computing power and data. We propose an efficient method with small model sizes and less training data with only 2-8 training instances per class. Our proposed method, AncSetFit, targets low data scenarios by anchoring the task and label information through sentence embeddings in fine-tuning a Sentence Transformer model. It uses contrastive learning and a triplet loss to enforce training instances of a class to be closest to its own textual semantic label information in the embedding space - and thereby learning to embed different class instances more distinct. AncSetFit obtains strong performance in data-sparse scenarios compared to existing methods across SST-5, Emotion detection, and AG News data, even with just two examples per class.", }

Few-shot classification is a powerful technique, but training requires substantial computing power and data. We propose an efficient method with small model sizes and less training data with only 2-8 training instances per class. Our proposed method, AncSetFit, targets low data scenarios by anchoring the task and label information through sentence embeddings in fine-tuning a Sentence Transformer model. It uses contrastive learning and a triplet loss to enforce training instances of a class to be closest to its own textual semantic label information in the embedding space - and thereby learning to embed different class instances more distinct. AncSetFit obtains strong performance in data-sparse scenarios compared to existing methods across SST-5, Emotion detection, and AG News data, even with just two examples per class.

Anchoring Fine-tuning of Sentence Transformer with Semantic Label Information for Efficient Truly Few-shot Classification

emnlp-main.692

Poster

https://aclanthology.org/2023.emnlp-main.693.bib

https://aclanthology.org/2023.emnlp-main.693/

@inproceedings{saad-falcon-etal-2023-udapdr, title = "{UDAPDR}: Unsupervised Domain Adaptation via {LLM} Prompting and Distillation of Rerankers", author = "Saad-Falcon, Jon and Khattab, Omar and Santhanam, Keshav and Florian, Radu and Franz, Martin and Roukos, Salim and Sil, Avirup and Sultan, Md and Potts, Christopher", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.693", doi = "10.18653/v1/2023.emnlp-main.693", pages = "11265--11279", abstract = "Many information retrieval tasks require large labeled datasets for fine-tuning. However, such datasets are often unavailable, and their utility for real-world applications can diminish quickly due to domain shifts. To address this challenge, we develop and motivate a method for using large language models (LLMs) to generate large numbers of synthetic queries cheaply. The method begins by generating a small number of synthetic queries using an expensive LLM. After that, a much less expensive one is used to create large numbers of synthetic queries, which are used to fine-tune a family of reranker models. These rerankers are then distilled into a single efficient retriever for use in the target domain. We show that this technique boosts zero-shot accuracy in long-tail domains and achieves substantially lower latency than standard reranking methods.", }

Many information retrieval tasks require large labeled datasets for fine-tuning. However, such datasets are often unavailable, and their utility for real-world applications can diminish quickly due to domain shifts. To address this challenge, we develop and motivate a method for using large language models (LLMs) to generate large numbers of synthetic queries cheaply. The method begins by generating a small number of synthetic queries using an expensive LLM. After that, a much less expensive one is used to create large numbers of synthetic queries, which are used to fine-tune a family of reranker models. These rerankers are then distilled into a single efficient retriever for use in the target domain. We show that this technique boosts zero-shot accuracy in long-tail domains and achieves substantially lower latency than standard reranking methods.

UDAPDR: Unsupervised Domain Adaptation via LLM Prompting and Distillation of Rerankers

emnlp-main.693

2303.00807

https://huggingface.co/papers/2303.00807

Poster

https://aclanthology.org/2023.emnlp-main.694.bib

https://aclanthology.org/2023.emnlp-main.694/

@inproceedings{hanley-durumeric-2023-tata, title = "{TATA}: Stance Detection via Topic-Agnostic and Topic-Aware Embeddings", author = "Hanley, Hans and Durumeric, Zakir", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.694", doi = "10.18653/v1/2023.emnlp-main.694", pages = "11280--11294", abstract = "Stance detection is important for understanding different attitudes and beliefs on the Internet. However, given that a passage{'}s stance toward a given topic is often highly dependent on that topic, building a stance detection model that generalizes to unseen topics is difficult. In this work, we propose using contrastive learning as well as an unlabeled dataset of news articles that cover a variety of different topics to train topic-agnostic/TAG and topic-aware/TAW embeddings for use in downstream stance detection. Combining these embeddings in our full TATA model, we achieve state-of-the-art performance across several public stance detection datasets (0.771 $F_1$-score on the Zero-shot VAST dataset). We release our code and data at https://github.com/hanshanley/tata.", }

Stance detection is important for understanding different attitudes and beliefs on the Internet. However, given that a passage{'}s stance toward a given topic is often highly dependent on that topic, building a stance detection model that generalizes to unseen topics is difficult. In this work, we propose using contrastive learning as well as an unlabeled dataset of news articles that cover a variety of different topics to train topic-agnostic/TAG and topic-aware/TAW embeddings for use in downstream stance detection. Combining these embeddings in our full TATA model, we achieve state-of-the-art performance across several public stance detection datasets (0.771 $F_1$-score on the Zero-shot VAST dataset). We release our code and data at https://github.com/hanshanley/tata.

TATA: Stance Detection via Topic-Agnostic and Topic-Aware Embeddings

emnlp-main.694

Poster

https://aclanthology.org/2023.emnlp-main.695.bib

https://aclanthology.org/2023.emnlp-main.695/

@inproceedings{yauney-etal-2023-data, title = "Data Similarity is Not Enough to Explain Language Model Performance", author = "Yauney, Gregory and Reif, Emily and Mimno, David", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.695", doi = "10.18653/v1/2023.emnlp-main.695", pages = "11295--11304", abstract = "Large language models achieve high performance on many but not all downstream tasks. The interaction between pretraining data and task data is commonly assumed to determine this variance: a task with data that is more similar to a model{'}s pretraining data is assumed to be easier for that model. We test whether distributional and example-specific similarity measures (embedding-, token- and model-based) correlate with language model performance through a large-scale comparison of the Pile and C4 pretraining datasets with downstream benchmarks. Similarity correlates with performance for multilingual datasets, but in other benchmarks, we surprisingly find that similarity metrics are not correlated with accuracy or even each other. This suggests that the relationship between pretraining data and downstream tasks is more complex than often assumed.", }

Large language models achieve high performance on many but not all downstream tasks. The interaction between pretraining data and task data is commonly assumed to determine this variance: a task with data that is more similar to a model{'}s pretraining data is assumed to be easier for that model. We test whether distributional and example-specific similarity measures (embedding-, token- and model-based) correlate with language model performance through a large-scale comparison of the Pile and C4 pretraining datasets with downstream benchmarks. Similarity correlates with performance for multilingual datasets, but in other benchmarks, we surprisingly find that similarity metrics are not correlated with accuracy or even each other. This suggests that the relationship between pretraining data and downstream tasks is more complex than often assumed.

Data Similarity is Not Enough to Explain Language Model Performance

emnlp-main.695

2311.09006

https://huggingface.co/papers/2311.09006

Poster

https://aclanthology.org/2023.emnlp-main.696.bib

https://aclanthology.org/2023.emnlp-main.696/

@inproceedings{zhu-etal-2023-zero, title = "Zero-shot Sharpness-Aware Quantization for Pre-trained Language Models", author = "Zhu, Miaoxi and Zhong, Qihuang and Shen, Li and Ding, Liang and Liu, Juhua and Du, Bo and Tao, Dacheng", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.696", doi = "10.18653/v1/2023.emnlp-main.696", pages = "11305--11327", abstract = "Quantization is a promising approach for reducing memory overhead and accelerating inference, especially in large pre-trained language model (PLM) scenarios. While having no access to original training data due to security and privacy concerns has emerged the demand for zero-shot quantization. Most of the cutting-edge zero-shot quantization methods primarily 1) apply to computer vision tasks, and 2) neglect of overfitting problem in the generative adversarial learning process, leading to sub-optimal performance. Motivated by this, we propose a novel zero-shot sharpness-aware quantization (ZSAQ) framework for the zero-shot quantization of various PLMs. The key algorithm in solving ZSAQ is the SAM-SGA optimization, which aims to improve the quantization accuracy and model generalization via optimizing a minimax problem. We theoretically prove the convergence rate for the minimax optimization problem and this result can be applied to other nonconvex-PL minimax optimization frameworks. Extensive experiments on 11 tasks demonstrate that our method brings consistent and significant performance gains on both discriminative and generative PLMs, i.e., up to +6.98 average score. Furthermore, we empirically validate that our method can effectively improve the model generalization.", }

Quantization is a promising approach for reducing memory overhead and accelerating inference, especially in large pre-trained language model (PLM) scenarios. While having no access to original training data due to security and privacy concerns has emerged the demand for zero-shot quantization. Most of the cutting-edge zero-shot quantization methods primarily 1) apply to computer vision tasks, and 2) neglect of overfitting problem in the generative adversarial learning process, leading to sub-optimal performance. Motivated by this, we propose a novel zero-shot sharpness-aware quantization (ZSAQ) framework for the zero-shot quantization of various PLMs. The key algorithm in solving ZSAQ is the SAM-SGA optimization, which aims to improve the quantization accuracy and model generalization via optimizing a minimax problem. We theoretically prove the convergence rate for the minimax optimization problem and this result can be applied to other nonconvex-PL minimax optimization frameworks. Extensive experiments on 11 tasks demonstrate that our method brings consistent and significant performance gains on both discriminative and generative PLMs, i.e., up to +6.98 average score. Furthermore, we empirically validate that our method can effectively improve the model generalization.

Zero-shot Sharpness-Aware Quantization for Pre-trained Language Models

emnlp-main.696

Poster

https://aclanthology.org/2023.emnlp-main.697.bib

https://aclanthology.org/2023.emnlp-main.697/

@inproceedings{ma-etal-2023-deciphering, title = "Deciphering Stereotypes in Pre-Trained Language Models", author = "Ma, Weicheng and Scheible, Henry and Wang, Brian and Veeramachaneni, Goutham and Chowdhary, Pratim and Sun, Alan and Koulogeorge, Andrew and Wang, Lili and Yang, Diyi and Vosoughi, Soroush", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.697", doi = "10.18653/v1/2023.emnlp-main.697", pages = "11328--11345", abstract = "Warning: This paper contains content that is stereotypical and may be upsetting. This paper addresses the issue of demographic stereotypes present in Transformer-based pre-trained language models (PLMs) and aims to deepen our understanding of how these biases are encoded in these models. To accomplish this, we introduce an easy-to-use framework for examining the stereotype-encoding behavior of PLMs through a combination of model probing and textual analyses. Our findings reveal that a small subset of attention heads within PLMs are primarily responsible for encoding stereotypes and that stereotypes toward specific minority groups can be identified using attention maps on these attention heads. Leveraging these insights, we propose an attention-head pruning method as a viable approach for debiasing PLMs, without compromising their language modeling capabilities or adversely affecting their performance on downstream tasks.", }

Warning: This paper contains content that is stereotypical and may be upsetting. This paper addresses the issue of demographic stereotypes present in Transformer-based pre-trained language models (PLMs) and aims to deepen our understanding of how these biases are encoded in these models. To accomplish this, we introduce an easy-to-use framework for examining the stereotype-encoding behavior of PLMs through a combination of model probing and textual analyses. Our findings reveal that a small subset of attention heads within PLMs are primarily responsible for encoding stereotypes and that stereotypes toward specific minority groups can be identified using attention maps on these attention heads. Leveraging these insights, we propose an attention-head pruning method as a viable approach for debiasing PLMs, without compromising their language modeling capabilities or adversely affecting their performance on downstream tasks.

Deciphering Stereotypes in Pre-Trained Language Models

emnlp-main.697

Oral

https://aclanthology.org/2023.emnlp-main.698.bib

https://aclanthology.org/2023.emnlp-main.698/

@inproceedings{cho-etal-2023-integrative, title = "An Integrative Survey on Mental Health Conversational Agents to Bridge Computer Science and Medical Perspectives", author = "Cho, Young Min and Rai, Sunny and Ungar, Lyle and Sedoc, Jo{\~a}o and Guntuku, Sharath", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.698", doi = "10.18653/v1/2023.emnlp-main.698", pages = "11346--11369", abstract = "Mental health conversational agents (a.k.a. chatbots) are widely studied for their potential to offer accessible support to those experiencing mental health challenges. Previous surveys on the topic primarily consider papers published in either computer science or medicine, leading to a divide in understanding and hindering the sharing of beneficial knowledge between both domains. To bridge this gap, we conduct a comprehensive literature review using the PRISMA framework, reviewing 534 papers published in both computer science and medicine. Our systematic review reveals 136 key papers on building mental health-related conversational agents with diverse characteristics of modeling and experimental design techniques. We find that computer science papers focus on LLM techniques and evaluating response quality using automated metrics with little attention to the application while medical papers use rule-based conversational agents and outcome metrics to measure the health outcomes of participants. Based on our findings on transparency, ethics, and cultural heterogeneity in this review, we provide a few recommendations to help bridge the disciplinary divide and enable the cross-disciplinary development of mental health conversational agents.", }

Mental health conversational agents (a.k.a. chatbots) are widely studied for their potential to offer accessible support to those experiencing mental health challenges. Previous surveys on the topic primarily consider papers published in either computer science or medicine, leading to a divide in understanding and hindering the sharing of beneficial knowledge between both domains. To bridge this gap, we conduct a comprehensive literature review using the PRISMA framework, reviewing 534 papers published in both computer science and medicine. Our systematic review reveals 136 key papers on building mental health-related conversational agents with diverse characteristics of modeling and experimental design techniques. We find that computer science papers focus on LLM techniques and evaluating response quality using automated metrics with little attention to the application while medical papers use rule-based conversational agents and outcome metrics to measure the health outcomes of participants. Based on our findings on transparency, ethics, and cultural heterogeneity in this review, we provide a few recommendations to help bridge the disciplinary divide and enable the cross-disciplinary development of mental health conversational agents.

An Integrative Survey on Mental Health Conversational Agents to Bridge Computer Science and Medical Perspectives

emnlp-main.698

2310.17017

Oral

https://aclanthology.org/2023.emnlp-main.699.bib

https://aclanthology.org/2023.emnlp-main.699/

@inproceedings{choi-etal-2023-llms, title = "Do {LLM}s Understand Social Knowledge? Evaluating the Sociability of Large Language Models with {S}oc{KET} Benchmark", author = "Choi, Minje and Pei, Jiaxin and Kumar, Sagar and Shu, Chang and Jurgens, David", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.699", doi = "10.18653/v1/2023.emnlp-main.699", pages = "11370--11403", abstract = "Large language models (LLMs) have been shown to perform well at a variety of syntactic, discourse, and reasoning tasks. While LLMs are increasingly deployed in many forms including conversational agents that interact with humans, we lack a grounded benchmark to measure how well LLMs understand social language. Here, we introduce a new theory-driven benchmark, SocKET, that contains 58 NLP tasks testing social knowledge which we group into five categories: humor {\&} sarcasm, offensiveness, sentiment {\&} emotion, and trustworthiness. In tests on the benchmark, we demonstrate that current models attain only moderate performance but reveal significant potential for task transfer among different types and categories of tasks, which were predicted from theory. Through zero-shot evaluations, we show that pretrained models already possess some innate but limited capabilities of social language understanding and training on one category of tasks can improve zero-shot testing on others. Our benchmark provides a systematic way to analyze model performance on an important dimension of language and points to clear room for improvement to build more socially-aware LLMs. The resources are released at https://github.com/minjechoi/SOCKET.", }

Large language models (LLMs) have been shown to perform well at a variety of syntactic, discourse, and reasoning tasks. While LLMs are increasingly deployed in many forms including conversational agents that interact with humans, we lack a grounded benchmark to measure how well LLMs understand social language. Here, we introduce a new theory-driven benchmark, SocKET, that contains 58 NLP tasks testing social knowledge which we group into five categories: humor {\&} sarcasm, offensiveness, sentiment {\&} emotion, and trustworthiness. In tests on the benchmark, we demonstrate that current models attain only moderate performance but reveal significant potential for task transfer among different types and categories of tasks, which were predicted from theory. Through zero-shot evaluations, we show that pretrained models already possess some innate but limited capabilities of social language understanding and training on one category of tasks can improve zero-shot testing on others. Our benchmark provides a systematic way to analyze model performance on an important dimension of language and points to clear room for improvement to build more socially-aware LLMs. The resources are released at https://github.com/minjechoi/SOCKET.

Do LLMs Understand Social Knowledge? Evaluating the Sociability of Large Language Models with SocKET Benchmark

emnlp-main.699

2305.14938

https://huggingface.co/papers/2305.14938

Oral

https://aclanthology.org/2023.emnlp-main.700.bib

https://aclanthology.org/2023.emnlp-main.700/

@inproceedings{yue-etal-2023-interventional, title = "Interventional Rationalization", author = "Yue, Linan and Liu, Qi and Wang, Li and An, Yanqing and Du, Yichao and Huang, Zhenya", editor = "Bouamor, Houda and Pino, Juan and Bali, Kalika", booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing", month = dec, year = "2023", address = "Singapore", publisher = "Association for Computational Linguistics", url = "https://aclanthology.org/2023.emnlp-main.700", doi = "10.18653/v1/2023.emnlp-main.700", pages = "11404--11418", abstract = "Selective rationalizations improve the explainability of neural networks by selecting a subsequence of the input (i.e., rationales) to explain the prediction results. Although existing methods have achieved promising results, they still suffer from adopting the spurious correlations in data (aka., shortcuts) to compose rationales and make predictions. Inspired by the causal theory, in this paper, we develop an interventional rationalization (Inter-RAT) to discover the causal rationales. Specifically, we first analyse the causalities among the input, rationales and results with a structural causal model. Then, we discover spurious correlations between the input and rationales, and between rationales and results, respectively, by identifying the confounder in the causalities. Next, based on the backdoor adjustment, we propose a causal intervention method to remove the spurious correlations between input and rationales. Further, we discuss reasons why spurious correlations between the selected rationales and results exist by analysing the limitations of the sparsity constraint in the rationalization, and employ the causal intervention method to remove these correlations. Extensive experimental results on three real-world datasets clearly validate the effectiveness of our proposed method. The source code of Inter-RAT is available at https://github.com/yuelinan/Codes-of-Inter-RAT.", }

Selective rationalizations improve the explainability of neural networks by selecting a subsequence of the input (i.e., rationales) to explain the prediction results. Although existing methods have achieved promising results, they still suffer from adopting the spurious correlations in data (aka., shortcuts) to compose rationales and make predictions. Inspired by the causal theory, in this paper, we develop an interventional rationalization (Inter-RAT) to discover the causal rationales. Specifically, we first analyse the causalities among the input, rationales and results with a structural causal model. Then, we discover spurious correlations between the input and rationales, and between rationales and results, respectively, by identifying the confounder in the causalities. Next, based on the backdoor adjustment, we propose a causal intervention method to remove the spurious correlations between input and rationales. Further, we discuss reasons why spurious correlations between the selected rationales and results exist by analysing the limitations of the sparsity constraint in the rationalization, and employ the causal intervention method to remove these correlations. Extensive experimental results on three real-world datasets clearly validate the effectiveness of our proposed method. The source code of Inter-RAT is available at https://github.com/yuelinan/Codes-of-Inter-RAT.

Interventional Rationalization

emnlp-main.700

Poster