bibtex_url
stringlengths 41
53
| acl_proceedings
stringlengths 38
50
| bibtext
stringlengths 528
3.02k
| abstract
stringlengths 17
2.35k
| authors
sequencelengths 1
44
| title
stringlengths 18
190
| id
stringlengths 7
19
| arxiv_id
stringlengths 10
10
⌀ | GitHub
sequencelengths 1
1
| paper_page
stringclasses 528
values | n_linked_authors
int64 -1
15
| upvotes
int64 -1
77
| num_comments
int64 -1
10
| n_authors
int64 -1
52
| Models
sequencelengths 0
100
| Datasets
sequencelengths 0
15
| Spaces
sequencelengths 0
46
| paper_page_exists_pre_conf
int64 0
1
| type
stringclasses 2
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
https://aclanthology.org/2023.emnlp-main.801.bib | https://aclanthology.org/2023.emnlp-main.801/ | @inproceedings{srivastava-etal-2023-mailex,
title = "{M}ail{E}x: Email Event and Argument Extraction",
author = "Srivastava, Saurabh and
Singh, Gaurav and
Matsumoto, Shou and
Raz, Ali and
Costa, Paulo and
Poore, Joshua and
Yao, Ziyu",
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.801",
doi = "10.18653/v1/2023.emnlp-main.801",
pages = "12964--12987",
abstract = "In this work, we present the first dataset, MailEx, for performing event extraction from conversational email threads. To this end, we first proposed a new taxonomy covering 10 event types and 76 arguments in the email domain. Our final dataset includes 1.5K email threads and {\textasciitilde}4K emails, which are annotated with a total of {\textasciitilde}8K event instances. To understand the task challenges, we conducted a series of experiments comparing three types of approaches, i.e., fine-tuned sequence labeling, fine-tuned generative extraction, and few-shot in-context learning. Our results showed that the task of email event extraction is far from being addressed, due to challenges lying in, e.g., extracting non-continuous, shared trigger spans, extracting non-named entity arguments, and modeling the email conversational history. Our work thus suggests more future investigations in this domain-specific event extraction task.",
}
| In this work, we present the first dataset, MailEx, for performing event extraction from conversational email threads. To this end, we first proposed a new taxonomy covering 10 event types and 76 arguments in the email domain. Our final dataset includes 1.5K email threads and {\textasciitilde}4K emails, which are annotated with a total of {\textasciitilde}8K event instances. To understand the task challenges, we conducted a series of experiments comparing three types of approaches, i.e., fine-tuned sequence labeling, fine-tuned generative extraction, and few-shot in-context learning. Our results showed that the task of email event extraction is far from being addressed, due to challenges lying in, e.g., extracting non-continuous, shared trigger spans, extracting non-named entity arguments, and modeling the email conversational history. Our work thus suggests more future investigations in this domain-specific event extraction task. | [
"Srivastava, Saurabh",
"Singh, Gaurav",
"Matsumoto, Shou",
"Raz, Ali",
"Costa, Paulo",
"Poore, Joshua",
"Yao, Ziyu"
] | MailEx: Email Event and Argument Extraction | emnlp-main.801 | 2305.13469 | [
"https://github.com/salokr/email-event-extraction"
] | https://huggingface.co/papers/2305.13469 | 0 | 0 | 0 | 7 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.802.bib | https://aclanthology.org/2023.emnlp-main.802/ | @inproceedings{wullach-chazan-2023-optimized,
title = "Optimized Tokenization for Transcribed Error Correction",
author = "Wullach, Tomer and
Chazan, Shlomo",
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.802",
doi = "10.18653/v1/2023.emnlp-main.802",
pages = "12988--12997",
abstract = "The challenges facing speech recognition systems, such as variations in pronunciations, adverse audio conditions, and the scarcity of labeled data, emphasize the necessity for a post-processing step that corrects recurring errors. Previous research has shown the advantages of employing dedicated error correction models, yet training such models requires large amounts of labeled data which is not easily obtained. To overcome this limitation, synthetic transcribed-like data is often utilized, however, bridging the distribution gap between transcribed errors and synthetic noise is not trivial. In this paper, we demonstrate that the performance of correction models can be significantly increased by training solely using synthetic data. Specifically, we empirically show that: (1) synthetic data generated using the error distribution derived from a set of transcribed data outperforms the common approach of applying random perturbations; (2) applying language-specific adjustments to the vocabulary of a BPE tokenizer strike a balance between adapting to unseen distributions and retaining knowledge of transcribed errors. We showcase the benefits of these key observations, and evaluate our approach using multiple languages, speech recognition systems and prominent speech recognition datasets.",
}
| The challenges facing speech recognition systems, such as variations in pronunciations, adverse audio conditions, and the scarcity of labeled data, emphasize the necessity for a post-processing step that corrects recurring errors. Previous research has shown the advantages of employing dedicated error correction models, yet training such models requires large amounts of labeled data which is not easily obtained. To overcome this limitation, synthetic transcribed-like data is often utilized, however, bridging the distribution gap between transcribed errors and synthetic noise is not trivial. In this paper, we demonstrate that the performance of correction models can be significantly increased by training solely using synthetic data. Specifically, we empirically show that: (1) synthetic data generated using the error distribution derived from a set of transcribed data outperforms the common approach of applying random perturbations; (2) applying language-specific adjustments to the vocabulary of a BPE tokenizer strike a balance between adapting to unseen distributions and retaining knowledge of transcribed errors. We showcase the benefits of these key observations, and evaluate our approach using multiple languages, speech recognition systems and prominent speech recognition datasets. | [
"Wullach, Tomer",
"Chazan, Shlomo"
] | Optimized Tokenization for Transcribed Error Correction | emnlp-main.802 | 2310.10704 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.803.bib | https://aclanthology.org/2023.emnlp-main.803/ | @inproceedings{su-etal-2023-beware,
title = "Beware of Model Collapse! Fast and Stable Test-time Adaptation for Robust Question Answering",
author = "Su, Yi and
Ji, Yixin and
Li, Juntao and
Ye, Hai 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.803",
doi = "10.18653/v1/2023.emnlp-main.803",
pages = "12998--13011",
abstract = "Although pre-trained language models (PLM) have achieved great success in question answering (QA), their robustness is still insufficient to support their practical applications, especially in the face of distribution shifts. Recently, test-time adaptation (TTA) has shown great potential for solving this problem, which adapts the model to fit the test samples at test time. However, TTA sometimes causes model collapse, making almost all the model outputs incorrect, which has raised concerns about its stability and reliability. In this paper, we delve into why TTA causes model collapse and find that the imbalanced label distribution inherent in QA is the reason for it. To address this problem, we propose Anti-Collapse Fast test-time adaptation (Anti-CF), which utilizes the source model{`}s output to regularize the update of the adapted model during test time. We further design an efficient side block to reduce its inference time. Extensive experiments on various distribution shift scenarios and pre-trained language models (e.g., XLM-RoBERTa, BLOOM) demonstrate that our method can achieve comparable or better results than previous TTA methods at a speed close to vanilla forward propagation, which is 1.8$\times$ to 4.4$\times$ speedup compared to previous TTA methods.",
}
| Although pre-trained language models (PLM) have achieved great success in question answering (QA), their robustness is still insufficient to support their practical applications, especially in the face of distribution shifts. Recently, test-time adaptation (TTA) has shown great potential for solving this problem, which adapts the model to fit the test samples at test time. However, TTA sometimes causes model collapse, making almost all the model outputs incorrect, which has raised concerns about its stability and reliability. In this paper, we delve into why TTA causes model collapse and find that the imbalanced label distribution inherent in QA is the reason for it. To address this problem, we propose Anti-Collapse Fast test-time adaptation (Anti-CF), which utilizes the source model{`}s output to regularize the update of the adapted model during test time. We further design an efficient side block to reduce its inference time. Extensive experiments on various distribution shift scenarios and pre-trained language models (e.g., XLM-RoBERTa, BLOOM) demonstrate that our method can achieve comparable or better results than previous TTA methods at a speed close to vanilla forward propagation, which is 1.8$\times$ to 4.4$\times$ speedup compared to previous TTA methods. | [
"Su, Yi",
"Ji, Yixin",
"Li, Juntao",
"Ye, Hai",
"Zhang, Min"
] | Beware of Model Collapse! Fast and Stable Test-time Adaptation for Robust Question Answering | emnlp-main.803 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.804.bib | https://aclanthology.org/2023.emnlp-main.804/ | @inproceedings{zhao-mao-2023-generative,
title = "Generative Adversarial Training with Perturbed Token Detection for Model Robustness",
author = "Zhao, Jiahao 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.804",
doi = "10.18653/v1/2023.emnlp-main.804",
pages = "13012--13025",
abstract = "Adversarial training is the dominant strategy towards model robustness. Current adversarial training methods typically apply perturbations to embedding representations, whereas actual text-based attacks introduce perturbations as discrete tokens. Thus there exists a gap between the continuous embedding representations and discrete text tokens that hampers the effectiveness of adversarial training. Moreover, the continuous representations of perturbations cannot be further utilized, resulting in the suboptimal performance. To bridge this gap for adversarial robustness, in this paper, we devise a novel generative adversarial training framework that integrates gradient-based learning, adversarial example generation and perturbed token detection. Our proposed framework consists of generative adversarial attack and adversarial training process. Specifically, in generative adversarial attack, the embeddings are shared between the classifier and the generative model, which enables the generative model to leverage the gradients from the classifier for generating perturbed tokens. Then, adversarial training process combines adversarial regularization with perturbed token detection to provide token-level supervision and improve the efficiency of sample utilization. Extensive experiments on five datasets from the AdvGLUE benchmark demonstrate that our framework significantly enhances the model robustness, surpassing the state-of-the-art results of ChatGPT by 10{\%} in average accuracy.",
}
| Adversarial training is the dominant strategy towards model robustness. Current adversarial training methods typically apply perturbations to embedding representations, whereas actual text-based attacks introduce perturbations as discrete tokens. Thus there exists a gap between the continuous embedding representations and discrete text tokens that hampers the effectiveness of adversarial training. Moreover, the continuous representations of perturbations cannot be further utilized, resulting in the suboptimal performance. To bridge this gap for adversarial robustness, in this paper, we devise a novel generative adversarial training framework that integrates gradient-based learning, adversarial example generation and perturbed token detection. Our proposed framework consists of generative adversarial attack and adversarial training process. Specifically, in generative adversarial attack, the embeddings are shared between the classifier and the generative model, which enables the generative model to leverage the gradients from the classifier for generating perturbed tokens. Then, adversarial training process combines adversarial regularization with perturbed token detection to provide token-level supervision and improve the efficiency of sample utilization. Extensive experiments on five datasets from the AdvGLUE benchmark demonstrate that our framework significantly enhances the model robustness, surpassing the state-of-the-art results of ChatGPT by 10{\%} in average accuracy. | [
"Zhao, Jiahao",
"Mao, Wenji"
] | Generative Adversarial Training with Perturbed Token Detection for Model Robustness | emnlp-main.804 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.805.bib | https://aclanthology.org/2023.emnlp-main.805/ | @inproceedings{park-caragea-2023-multi,
title = "Multi-Task Knowledge Distillation with Embedding Constraints for Scholarly Keyphrase Boundary Classification",
author = "Park, Seo and
Caragea, Cornelia",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.emnlp-main.805",
doi = "10.18653/v1/2023.emnlp-main.805",
pages = "13026--13042",
abstract = "The task of scholarly keyphrase boundary classification aims at identifying keyphrases from scientific papers and classifying them with their types from a set of predefined classes (e.g., task, process, or material). Despite the importance of keyphrases and their types in many downstream applications including indexing, searching, and question answering over scientific documents, scholarly keyphrase boundary classification is still an under-explored task. In this work, we propose a novel embedding constraint on multi-task knowledge distillation which enforces the teachers (single-task models) and the student (multi-task model) similarity in the embedding space. Specifically, we enforce that the student model is trained not only to imitate the teachers{'} output distribution over classes, but also to produce language representations that are similar to those produced by the teachers. Our results show that the proposed approach outperforms previous works and strong baselines on three datasets of scientific documents.",
}
| The task of scholarly keyphrase boundary classification aims at identifying keyphrases from scientific papers and classifying them with their types from a set of predefined classes (e.g., task, process, or material). Despite the importance of keyphrases and their types in many downstream applications including indexing, searching, and question answering over scientific documents, scholarly keyphrase boundary classification is still an under-explored task. In this work, we propose a novel embedding constraint on multi-task knowledge distillation which enforces the teachers (single-task models) and the student (multi-task model) similarity in the embedding space. Specifically, we enforce that the student model is trained not only to imitate the teachers{'} output distribution over classes, but also to produce language representations that are similar to those produced by the teachers. Our results show that the proposed approach outperforms previous works and strong baselines on three datasets of scientific documents. | [
"Park, Seo",
"Caragea, Cornelia"
] | Multi-Task Knowledge Distillation with Embedding Constraints for Scholarly Keyphrase Boundary Classification | emnlp-main.805 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.806.bib | https://aclanthology.org/2023.emnlp-main.806/ | @inproceedings{li-etal-2023-set,
title = "Set Learning for Generative Information Extraction",
author = "Li, Jiangnan and
Zhang, Yice and
Liang, Bin and
Wong, Kam-Fai and
Xu, Ruifeng",
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.806",
doi = "10.18653/v1/2023.emnlp-main.806",
pages = "13043--13052",
abstract = "Recent efforts have endeavored to employ the sequence-to-sequence (Seq2Seq) model in Information Extraction (IE) due to its potential to tackle multiple IE tasks in a unified manner. Under this formalization, multiple structured objects are concatenated as the target sequence in a predefined order. However, structured objects, by their nature, constitute an unordered set. Consequently, this formalization introduces a potential order bias, which can impair model learning. Targeting this issue, this paper proposes a set learning approach that considers multiple permutations of structured objects to optimize set probability approximately. Notably, our approach does not require any modifications to model structures, making it easily integrated into existing generative IE frameworks. Experiments show that our method consistently improves existing frameworks on vast tasks and datasets.",
}
| Recent efforts have endeavored to employ the sequence-to-sequence (Seq2Seq) model in Information Extraction (IE) due to its potential to tackle multiple IE tasks in a unified manner. Under this formalization, multiple structured objects are concatenated as the target sequence in a predefined order. However, structured objects, by their nature, constitute an unordered set. Consequently, this formalization introduces a potential order bias, which can impair model learning. Targeting this issue, this paper proposes a set learning approach that considers multiple permutations of structured objects to optimize set probability approximately. Notably, our approach does not require any modifications to model structures, making it easily integrated into existing generative IE frameworks. Experiments show that our method consistently improves existing frameworks on vast tasks and datasets. | [
"Li, Jiangnan",
"Zhang, Yice",
"Liang, Bin",
"Wong, Kam-Fai",
"Xu, Ruifeng"
] | Set Learning for Generative Information Extraction | emnlp-main.806 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.807.bib | https://aclanthology.org/2023.emnlp-main.807/ | @inproceedings{kritharoula-etal-2023-large,
title = "Large Language Models and Multimodal Retrieval for Visual Word Sense Disambiguation",
author = "Kritharoula, Anastasia and
Lymperaiou, Maria and
Stamou, Giorgos",
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.807",
doi = "10.18653/v1/2023.emnlp-main.807",
pages = "13053--13077",
abstract = "Visual Word Sense Disambiguation (VWSD) is a novel challenging task with the goal of retrieving an image among a set of candidates, which better represents the meaning of an ambiguous word within a given context. In this paper, we make a substantial step towards unveiling this interesting task by applying a varying set of approaches. Since VWSD is primarily a text-image retrieval task, we explore the latest transformer-based methods for multimodal retrieval. Additionally, we utilize Large Language Models (LLMs) as knowledge bases to enhance the given phrases and resolve ambiguity related to the target word. We also study VWSD as a unimodal problem by converting to text-to-text and image-to-image retrieval, as well as question-answering (QA), to fully explore the capabilities of relevant models. To tap into the implicit knowledge of LLMs, we experiment with Chain-of-Thought (CoT) prompting to guide explainable answer generation. On top of all, we train a learn to rank (LTR) model in order to combine our different modules, achieving competitive ranking results. Extensive experiments on VWSD demonstrate valuable insights to effectively drive future directions.",
}
| Visual Word Sense Disambiguation (VWSD) is a novel challenging task with the goal of retrieving an image among a set of candidates, which better represents the meaning of an ambiguous word within a given context. In this paper, we make a substantial step towards unveiling this interesting task by applying a varying set of approaches. Since VWSD is primarily a text-image retrieval task, we explore the latest transformer-based methods for multimodal retrieval. Additionally, we utilize Large Language Models (LLMs) as knowledge bases to enhance the given phrases and resolve ambiguity related to the target word. We also study VWSD as a unimodal problem by converting to text-to-text and image-to-image retrieval, as well as question-answering (QA), to fully explore the capabilities of relevant models. To tap into the implicit knowledge of LLMs, we experiment with Chain-of-Thought (CoT) prompting to guide explainable answer generation. On top of all, we train a learn to rank (LTR) model in order to combine our different modules, achieving competitive ranking results. Extensive experiments on VWSD demonstrate valuable insights to effectively drive future directions. | [
"Kritharoula, Anastasia",
"Lymperaiou, Maria",
"Stamou, Giorgos"
] | Large Language Models and Multimodal Retrieval for Visual Word Sense Disambiguation | emnlp-main.807 | 2310.14025 | [
"https://github.com/anastasiakrith/multimodal-retrieval-for-vwsd"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Oral |
|
https://aclanthology.org/2023.emnlp-main.808.bib | https://aclanthology.org/2023.emnlp-main.808/ | @inproceedings{chawla-etal-2023-selfish,
title = "Be Selfish, But Wisely: Investigating the Impact of Agent Personality in Mixed-Motive Human-Agent Interactions",
author = "Chawla, Kushal and
Wu, Ian and
Rong, Yu and
Lucas, Gale and
Gratch, Jonathan",
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.808",
doi = "10.18653/v1/2023.emnlp-main.808",
pages = "13078--13092",
abstract = "A natural way to design a negotiation dialogue system is via self-play RL: train an agent that learns to maximize its performance by interacting with a simulated user that has been designed to imitate human-human dialogue data. Although this procedure has been adopted in prior work, we find that it results in a fundamentally flawed system that fails to learn the value of compromise in a negotiation, which can often lead to no agreements (i.e., the partner walking away without a deal), ultimately hurting the model{'}s overall performance. We investigate this observation in the context of DealOrNoDeal task, a multi-issue negotiation over books, hats, and balls. Grounded in negotiation theory from Economics, we modify the training procedure in two novel ways to design agents with diverse personalities and analyze their performance with human partners. We find that although both techniques show promise, a selfish agent, which maximizes its own performance while also avoiding walkaways, performs superior to other variants by implicitly learning to generate value for both itself and the negotiation partner. We discuss the implications of our findings for what it means to be a successful negotiation dialogue system and how these systems should be designed in the future.",
}
| A natural way to design a negotiation dialogue system is via self-play RL: train an agent that learns to maximize its performance by interacting with a simulated user that has been designed to imitate human-human dialogue data. Although this procedure has been adopted in prior work, we find that it results in a fundamentally flawed system that fails to learn the value of compromise in a negotiation, which can often lead to no agreements (i.e., the partner walking away without a deal), ultimately hurting the model{'}s overall performance. We investigate this observation in the context of DealOrNoDeal task, a multi-issue negotiation over books, hats, and balls. Grounded in negotiation theory from Economics, we modify the training procedure in two novel ways to design agents with diverse personalities and analyze their performance with human partners. We find that although both techniques show promise, a selfish agent, which maximizes its own performance while also avoiding walkaways, performs superior to other variants by implicitly learning to generate value for both itself and the negotiation partner. We discuss the implications of our findings for what it means to be a successful negotiation dialogue system and how these systems should be designed in the future. | [
"Chawla, Kushal",
"Wu, Ian",
"Rong, Yu",
"Lucas, Gale",
"Gratch, Jonathan"
] | Be Selfish, But Wisely: Investigating the Impact of Agent Personality in Mixed-Motive Human-Agent Interactions | emnlp-main.808 | 2310.14404 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.809.bib | https://aclanthology.org/2023.emnlp-main.809/ | @inproceedings{diao-etal-2023-doolittle,
title = "Doolittle: Benchmarks and Corpora for Academic Writing Formalization",
author = "Diao, Shizhe and
Lei, Yongyu and
Pan, Liangming and
Fang, Tianqing and
Zhou, Wangchunshu and
Keh, Sedrick and
Kan, Min-Yen and
Zhang, Tong",
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.809",
doi = "10.18653/v1/2023.emnlp-main.809",
pages = "13093--13111",
abstract = "Improving the quality of academic writing is a meaningful but challenging task. Conventional methods of language refinement focus on narrow, specific linguistic features within isolated sentences, such as grammatical errors and improper word use. We propose a more general task, Academic Writing Formalization (AWF), to improve the overall quality of formal academic writing at the paragraph level. We formulate this language refinement task as a formal text style transfer task which transfers informal-academic text to formal-academic and contribute a large-scale non-parallel dataset, Doolittle, for this purpose. Concurrently, we apply a method named metric-oriented reinforcement learning (MORL) to two large language models (LLM) where we incorporate different levels of automatic feedback into the training process. Our experiments reveal that existing text transfer models and grammatical error correction models address certain aspects of AWF but still have a significant performance gap compared to human performance. Meanwhile, language models fine-tuned with our MORL method exhibit considerably improved performance, rivaling the latest chatbot ChatGPT, but still have a non-negligible gap compared to the ground truth formal-academic texts in Doolittle.",
}
| Improving the quality of academic writing is a meaningful but challenging task. Conventional methods of language refinement focus on narrow, specific linguistic features within isolated sentences, such as grammatical errors and improper word use. We propose a more general task, Academic Writing Formalization (AWF), to improve the overall quality of formal academic writing at the paragraph level. We formulate this language refinement task as a formal text style transfer task which transfers informal-academic text to formal-academic and contribute a large-scale non-parallel dataset, Doolittle, for this purpose. Concurrently, we apply a method named metric-oriented reinforcement learning (MORL) to two large language models (LLM) where we incorporate different levels of automatic feedback into the training process. Our experiments reveal that existing text transfer models and grammatical error correction models address certain aspects of AWF but still have a significant performance gap compared to human performance. Meanwhile, language models fine-tuned with our MORL method exhibit considerably improved performance, rivaling the latest chatbot ChatGPT, but still have a non-negligible gap compared to the ground truth formal-academic texts in Doolittle. | [
"Diao, Shizhe",
"Lei, Yongyu",
"Pan, Liangming",
"Fang, Tianqing",
"Zhou, Wangchunshu",
"Keh, Sedrick",
"Kan, Min-Yen",
"Zhang, Tong"
] | Doolittle: Benchmarks and Corpora for Academic Writing Formalization | emnlp-main.809 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.810.bib | https://aclanthology.org/2023.emnlp-main.810/ | @inproceedings{chen-etal-2023-token,
title = "Token Prediction as Implicit Classification to Identify {LLM}-Generated Text",
author = "Chen, Yutian and
Kang, Hao and
Zhai, Vivian and
Li, Liangze and
Singh, Rita and
Raj, Bhiksha",
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.810",
doi = "10.18653/v1/2023.emnlp-main.810",
pages = "13112--13120",
abstract = "This paper introduces a novel approach for identifying the possible large language models (LLMs) involved in text generation. Instead of adding an additional classification layer to a base LM, we reframe the classification task as a next-token prediction task and directly fine-tune the base LM to perform it. We utilize the Text-to-Text Transfer Transformer (T5) model as the backbone for our experiments. We compared our approach to the more direct approach of utilizing hidden states for classification. Evaluation shows the exceptional performance of our method in the text classification task, highlighting its simplicity and efficiency. Furthermore, interpretability studies on the features extracted by our model reveal its ability to differentiate distinctive writing styles among various LLMs even in the absence of an explicit classifier. We also collected a dataset named OpenLLMText, containing approximately 340k text samples from human and LLMs, including GPT3.5, PaLM, LLaMA, and GPT2.",
}
| This paper introduces a novel approach for identifying the possible large language models (LLMs) involved in text generation. Instead of adding an additional classification layer to a base LM, we reframe the classification task as a next-token prediction task and directly fine-tune the base LM to perform it. We utilize the Text-to-Text Transfer Transformer (T5) model as the backbone for our experiments. We compared our approach to the more direct approach of utilizing hidden states for classification. Evaluation shows the exceptional performance of our method in the text classification task, highlighting its simplicity and efficiency. Furthermore, interpretability studies on the features extracted by our model reveal its ability to differentiate distinctive writing styles among various LLMs even in the absence of an explicit classifier. We also collected a dataset named OpenLLMText, containing approximately 340k text samples from human and LLMs, including GPT3.5, PaLM, LLaMA, and GPT2. | [
"Chen, Yutian",
"Kang, Hao",
"Zhai, Vivian",
"Li, Liangze",
"Singh, Rita",
"Raj, Bhiksha"
] | Token Prediction as Implicit Classification to Identify LLM-Generated Text | emnlp-main.810 | 2311.08723 | [
"https://github.com/markchenyutian/t5-sentinel-public"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.811.bib | https://aclanthology.org/2023.emnlp-main.811/ | @inproceedings{akter-etal-2023-evaluation,
title = "On Evaluation of {B}angla Word Analogies",
author = "Akter, Mousumi and
Sarkar, Souvika and
Karmaker Santu, Shubhra Kanti",
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.811",
doi = "10.18653/v1/2023.emnlp-main.811",
pages = "13121--13127",
abstract = "This paper presents a benchmark dataset of Bangla word analogies for evaluating the quality of existing Bangla word embeddings. Despite being the 7th largest spoken language in the world, Bangla is still a low-resource language and popular NLP models often struggle to perform well on Bangla data sets. Therefore, developing a robust evaluation set is crucial for benchmarking and guiding future research on improving Bangla word embeddings, which is currently missing. To address this issue, we introduce a new evaluation set of 16,678 unique word analogies in Bangla as well as a translated and curated version of the original Mikolov dataset (10,594 samples) in Bangla. Our experiments with different state-of-the-art embedding models reveal that current Bangla word embeddings struggle to achieve high accuracy on both data sets, demonstrating a significant gap in multilingual NLP research.",
}
| This paper presents a benchmark dataset of Bangla word analogies for evaluating the quality of existing Bangla word embeddings. Despite being the 7th largest spoken language in the world, Bangla is still a low-resource language and popular NLP models often struggle to perform well on Bangla data sets. Therefore, developing a robust evaluation set is crucial for benchmarking and guiding future research on improving Bangla word embeddings, which is currently missing. To address this issue, we introduce a new evaluation set of 16,678 unique word analogies in Bangla as well as a translated and curated version of the original Mikolov dataset (10,594 samples) in Bangla. Our experiments with different state-of-the-art embedding models reveal that current Bangla word embeddings struggle to achieve high accuracy on both data sets, demonstrating a significant gap in multilingual NLP research. | [
"Akter, Mousumi",
"Sarkar, Souvika",
"Karmaker Santu, Shubhra Kanti"
] | On Evaluation of Bangla Word Analogies | emnlp-main.811 | 2304.04613 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.812.bib | https://aclanthology.org/2023.emnlp-main.812/ | @inproceedings{tan-etal-2023-reconstruct,
title = "Reconstruct Before Summarize: An Efficient Two-Step Framework for Condensing and Summarizing Meeting Transcripts",
author = "Tan, Haochen and
Wu, Han and
Shao, Wei and
Zhang, Xinyun and
Zhan, Mingjie and
Hou, Zhaohui and
Liang, Ding and
Song, Linqi",
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.812",
doi = "10.18653/v1/2023.emnlp-main.812",
pages = "13128--13141",
abstract = "Meetings typically involve multiple participants and lengthy conversations, resulting in redundant and trivial content. To overcome these challenges, we propose a two-step framework, Reconstruct before Summarize (RbS), for effective and efficient meeting summarization. RbS first leverages a self-supervised paradigm to annotate essential contents by reconstructing the meeting transcripts. Secondly, we propose a relative positional bucketing (RPB) algorithm to equip (conventional) summarization models to generate the summary. Despite the additional reconstruction process, our proposed RPB significantly compresses the input, leading to faster processing and reduced memory consumption compared to traditional summarization methods. We validate the effectiveness and efficiency of our method through extensive evaluations and analyses. On two meeting summarization datasets, AMI and ICSI, our approach outperforms previous state-of-the-art approaches without relying on large-scale pre-training or expert-grade annotating tools.",
}
| Meetings typically involve multiple participants and lengthy conversations, resulting in redundant and trivial content. To overcome these challenges, we propose a two-step framework, Reconstruct before Summarize (RbS), for effective and efficient meeting summarization. RbS first leverages a self-supervised paradigm to annotate essential contents by reconstructing the meeting transcripts. Secondly, we propose a relative positional bucketing (RPB) algorithm to equip (conventional) summarization models to generate the summary. Despite the additional reconstruction process, our proposed RPB significantly compresses the input, leading to faster processing and reduced memory consumption compared to traditional summarization methods. We validate the effectiveness and efficiency of our method through extensive evaluations and analyses. On two meeting summarization datasets, AMI and ICSI, our approach outperforms previous state-of-the-art approaches without relying on large-scale pre-training or expert-grade annotating tools. | [
"Tan, Haochen",
"Wu, Han",
"Shao, Wei",
"Zhang, Xinyun",
"Zhan, Mingjie",
"Hou, Zhaohui",
"Liang, Ding",
"Song, Linqi"
] | Reconstruct Before Summarize: An Efficient Two-Step Framework for Condensing and Summarizing Meeting Transcripts | emnlp-main.812 | 2305.07988 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.813.bib | https://aclanthology.org/2023.emnlp-main.813/ | @inproceedings{liang-etal-2023-xlm,
title = "{XLM}-{V}: Overcoming the Vocabulary Bottleneck in Multilingual Masked Language Models",
author = "Liang, Davis and
Gonen, Hila and
Mao, Yuning and
Hou, Rui and
Goyal, Naman and
Ghazvininejad, Marjan and
Zettlemoyer, Luke and
Khabsa, Madian",
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.813",
doi = "10.18653/v1/2023.emnlp-main.813",
pages = "13142--13152",
abstract = "Large multilingual language models typically rely on a single vocabulary shared across 100+ languages. As these models have increased in parameter count and depth, vocabulary size has remained largely unchanged. This \textit{vocabulary bottleneck} limits the representational capabilities of multilingual models like XLM-R. In this paper, we introduce a new approach for scaling to very large multilingual vocabularies by de-emphasizing token sharing between languages with little lexical overlap and assigning vocabulary capacity to achieve sufficient coverage for each individual language. Tokenizations using our vocabulary are typically more semantically meaningful and shorter compared to XLM-R. Leveraging this improved vocabulary, we train XLM-V, a multilingual language model with a one million token vocabulary. XLM-V outperforms XLM-R on every task we tested on ranging from natural language inference (XNLI), question answering (MLQA, XQuAD, TyDiQA), to named entity recognition (WikiAnn). XLM-V is particularly effective on low-resource language tasks and outperforms XLM-R by 11.2{\%} and 5.8{\%} absolute on MasakhaNER and Americas NLI, respectively.",
}
| Large multilingual language models typically rely on a single vocabulary shared across 100+ languages. As these models have increased in parameter count and depth, vocabulary size has remained largely unchanged. This \textit{vocabulary bottleneck} limits the representational capabilities of multilingual models like XLM-R. In this paper, we introduce a new approach for scaling to very large multilingual vocabularies by de-emphasizing token sharing between languages with little lexical overlap and assigning vocabulary capacity to achieve sufficient coverage for each individual language. Tokenizations using our vocabulary are typically more semantically meaningful and shorter compared to XLM-R. Leveraging this improved vocabulary, we train XLM-V, a multilingual language model with a one million token vocabulary. XLM-V outperforms XLM-R on every task we tested on ranging from natural language inference (XNLI), question answering (MLQA, XQuAD, TyDiQA), to named entity recognition (WikiAnn). XLM-V is particularly effective on low-resource language tasks and outperforms XLM-R by 11.2{\%} and 5.8{\%} absolute on MasakhaNER and Americas NLI, respectively. | [
"Liang, Davis",
"Gonen, Hila",
"Mao, Yuning",
"Hou, Rui",
"Goyal, Naman",
"Ghazvininejad, Marjan",
"Zettlemoyer, Luke",
"Khabsa, Madian"
] | XLM-V: Overcoming the Vocabulary Bottleneck in Multilingual Masked Language Models | emnlp-main.813 | 2301.10472 | [
"https://huggingface.co/facebook/xlm-v-base"
] | https://huggingface.co/papers/2301.10472 | 2 | 0 | 0 | 8 | [
"facebook/xlm-v-base",
"MoritzLaurer/xlm-v-base-mnli-xnli",
"orgcatorg/xlm-v-base-ner",
"imvladikon/xlmv_base"
] | [] | [
"awacke1/HEDIS.Dash.Component.Top.Clinical.Terminology.Vocabulary",
"mmaguero/Auto-Complete_Semantic",
"annalyzin/chinese_sentence_grading"
] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.814.bib | https://aclanthology.org/2023.emnlp-main.814/ | @inproceedings{shao-etal-2023-character,
title = "Character-{LLM}: A Trainable Agent for Role-Playing",
author = "Shao, Yunfan and
Li, Linyang and
Dai, Junqi and
Qiu, Xipeng",
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.814",
pages = "13153--13187",
abstract = "Large language models (LLMs) can be used to serve as agents to simulate human behaviors, given the powerful ability to understand human instructions and provide high-quality generated texts. Such ability stimulates us to wonder whether LLMs can simulate a person in a higher form than simple human behaviors. Therefore, we aim to train an agent with the profile, experience, and emotional states of a specific person instead of using limited prompts to instruct ChatGPT API. In this work, we introduce Character-LLM that teach LLMs to act as specific people such as Beethoven, Queen Cleopatra, Julius Caesar, etc. Our method focuses on editing profiles as experiences of a certain character and training models to be personal simulacra with these experiences. To assess the effectiveness of our approach, we build a test playground that interviews trained agents and evaluates whether the agents \textit{memorize} their characters and experiences. Experimental results show interesting observations that help build future simulacra of humankind.",
}
| Large language models (LLMs) can be used to serve as agents to simulate human behaviors, given the powerful ability to understand human instructions and provide high-quality generated texts. Such ability stimulates us to wonder whether LLMs can simulate a person in a higher form than simple human behaviors. Therefore, we aim to train an agent with the profile, experience, and emotional states of a specific person instead of using limited prompts to instruct ChatGPT API. In this work, we introduce Character-LLM that teach LLMs to act as specific people such as Beethoven, Queen Cleopatra, Julius Caesar, etc. Our method focuses on editing profiles as experiences of a certain character and training models to be personal simulacra with these experiences. To assess the effectiveness of our approach, we build a test playground that interviews trained agents and evaluates whether the agents \textit{memorize} their characters and experiences. Experimental results show interesting observations that help build future simulacra of humankind. | [
"Shao, Yunfan",
"Li, Linyang",
"Dai, Junqi",
"Qiu, Xipeng"
] | Character-LLM: A Trainable Agent for Role-Playing | emnlp-main.814 | 2310.10158 | [
"https://github.com/choosewhatulike/trainable-agents"
] | https://huggingface.co/papers/2310.10158 | 2 | 1 | 0 | 4 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.815.bib | https://aclanthology.org/2023.emnlp-main.815/ | @inproceedings{hoyle-etal-2023-natural,
title = "Natural Language Decompositions of Implicit Content Enable Better Text Representations",
author = "Hoyle, Alexander and
Sarkar, Rupak and
Goel, Pranav and
Resnik, Philip",
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.815",
doi = "10.18653/v1/2023.emnlp-main.815",
pages = "13188--13214",
abstract = "When people interpret text, they rely on inferences that go beyond the observed language itself. Inspired by this observation, we introduce a method for the analysis of text that takes implicitly communicated content explicitly into account. We use a large language model to produce sets of propositions that are inferentially related to the text that has been observed, then validate the plausibility of the generated content via human judgments. Incorporating these explicit representations of implicit content proves useful in multiple problem settings that involve the human interpretation of utterances: assessing the similarity of arguments, making sense of a body of opinion data, and modeling legislative behavior. Our results suggest that modeling the meanings behind observed language, rather than the literal text alone, is a valuable direction for NLP and particularly its applications to social science.",
}
| When people interpret text, they rely on inferences that go beyond the observed language itself. Inspired by this observation, we introduce a method for the analysis of text that takes implicitly communicated content explicitly into account. We use a large language model to produce sets of propositions that are inferentially related to the text that has been observed, then validate the plausibility of the generated content via human judgments. Incorporating these explicit representations of implicit content proves useful in multiple problem settings that involve the human interpretation of utterances: assessing the similarity of arguments, making sense of a body of opinion data, and modeling legislative behavior. Our results suggest that modeling the meanings behind observed language, rather than the literal text alone, is a valuable direction for NLP and particularly its applications to social science. | [
"Hoyle, Alex",
"er",
"Sarkar, Rupak",
"Goel, Pranav",
"Resnik, Philip"
] | Natural Language Decompositions of Implicit Content Enable Better Text Representations | emnlp-main.815 | 2305.14583 | [
"https://github.com/ahoho/inferential-decompositions"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.816.bib | https://aclanthology.org/2023.emnlp-main.816/ | @inproceedings{wang-etal-2023-scalable,
title = "A Scalable Framework for Table of Contents Extraction from Complex {ESG} Annual Reports",
author = "Wang, Xinyu and
Gui, Lin and
He, Yulan",
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.816",
doi = "10.18653/v1/2023.emnlp-main.816",
pages = "13215--13229",
abstract = "Table of contents (ToC) extraction centres on structuring documents in a hierarchical manner. In this paper, we propose a new dataset, ESGDoc, comprising 1,093 ESG annual reports from 563 companies spanning from 2001 to 2022. These reports pose significant challenges due to their diverse structures and extensive length. To address these challenges, we propose a new framework for Toc extraction, consisting of three steps: (1) Constructing an initial tree of text blocks based on reading order and font sizes; (2) Modelling each tree node (or text block) independently by considering its contextual information captured in node-centric subtree; (3) Modifying the original tree by taking appropriate action on each tree node (Keep, Delete, or Move). This construction-modelling-modification (CMM) process offers several benefits. It eliminates the need for pairwise modelling of section headings as in previous approaches, making document segmentation practically feasible. By incorporating structured information, each section heading can leverage both local and long-distance context relevant to itself. Experimental results show that our approach outperforms the previous state-of-the-art baseline with a fraction of running time. Our framework proves its scalability by effectively handling documents of any length.",
}
| Table of contents (ToC) extraction centres on structuring documents in a hierarchical manner. In this paper, we propose a new dataset, ESGDoc, comprising 1,093 ESG annual reports from 563 companies spanning from 2001 to 2022. These reports pose significant challenges due to their diverse structures and extensive length. To address these challenges, we propose a new framework for Toc extraction, consisting of three steps: (1) Constructing an initial tree of text blocks based on reading order and font sizes; (2) Modelling each tree node (or text block) independently by considering its contextual information captured in node-centric subtree; (3) Modifying the original tree by taking appropriate action on each tree node (Keep, Delete, or Move). This construction-modelling-modification (CMM) process offers several benefits. It eliminates the need for pairwise modelling of section headings as in previous approaches, making document segmentation practically feasible. By incorporating structured information, each section heading can leverage both local and long-distance context relevant to itself. Experimental results show that our approach outperforms the previous state-of-the-art baseline with a fraction of running time. Our framework proves its scalability by effectively handling documents of any length. | [
"Wang, Xinyu",
"Gui, Lin",
"He, Yulan"
] | A Scalable Framework for Table of Contents Extraction from Complex ESG Annual Reports | emnlp-main.816 | 2310.18073 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.817.bib | https://aclanthology.org/2023.emnlp-main.817/ | @inproceedings{zhang-etal-2023-semantic,
title = "Semantic Space Grounded Weighted Decoding for Multi-Attribute Controllable Dialogue Generation",
author = "Zhang, Zhiling and
Wu, Mengyue 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.817",
doi = "10.18653/v1/2023.emnlp-main.817",
pages = "13230--13243",
abstract = "Controlling chatbot utterance generation with multiple attributes such as personalities, emotions and dialogue acts is a practically useful but under-studied problem. We propose a novel framework called DASC that possesses strong controllability with a weighted decoding paradigm, while improving generation quality with the grounding in an attribute semantics space. Generation with multiple attributes is then intuitively implemented with an interpolation of multiple attribute embeddings, which results in substantial reduction in the model sizes. Experiments show that DASC can achieve high control accuracy in generation task with the simultaneous control of 3 aspects while also producing interesting and reasonably sensible responses, even in an out-of-distribution robustness test.",
}
| Controlling chatbot utterance generation with multiple attributes such as personalities, emotions and dialogue acts is a practically useful but under-studied problem. We propose a novel framework called DASC that possesses strong controllability with a weighted decoding paradigm, while improving generation quality with the grounding in an attribute semantics space. Generation with multiple attributes is then intuitively implemented with an interpolation of multiple attribute embeddings, which results in substantial reduction in the model sizes. Experiments show that DASC can achieve high control accuracy in generation task with the simultaneous control of 3 aspects while also producing interesting and reasonably sensible responses, even in an out-of-distribution robustness test. | [
"Zhang, Zhiling",
"Wu, Mengyue",
"Zhu, Kenny"
] | Semantic Space Grounded Weighted Decoding for Multi-Attribute Controllable Dialogue Generation | emnlp-main.817 | 2305.02820 | [
"https://github.com/blmoistawinde/dasc"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.818.bib | https://aclanthology.org/2023.emnlp-main.818/ | @inproceedings{choenni-etal-2023-languages,
title = "How do languages influence each other? Studying cross-lingual data sharing during {LM} fine-tuning",
author = "Choenni, Rochelle and
Garrette, Dan and
Shutova, Ekaterina",
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.818",
doi = "10.18653/v1/2023.emnlp-main.818",
pages = "13244--13257",
abstract = "Multilingual language models (MLMs) are jointly trained on data from many different languages such that representation of individual languages can benefit from other languages{'} data. Impressive performance in zero-shot cross-lingual transfer shows that these models are able to exploit this property. Yet, it remains unclear to what extent, and under which conditions, languages rely on each other{'}s data. To answer this question, we use TracIn (Pruthi et al., 2020), a training data attribution (TDA) method, to retrieve training samples from multilingual data that are most influential for test predictions in a given language. This allows us to analyse cross-lingual sharing mechanisms of MLMs from a new perspective. While previous work studied cross-lingual sharing at the model parameter level, we present the first approach to study it at the data level. We find that MLMs rely on data from multiple languages during fine-tuning and this reliance increases as fine-tuning progresses. We further find that training samples from other languages can both reinforce and complement the knowledge acquired from data of the test language itself.",
}
| Multilingual language models (MLMs) are jointly trained on data from many different languages such that representation of individual languages can benefit from other languages{'} data. Impressive performance in zero-shot cross-lingual transfer shows that these models are able to exploit this property. Yet, it remains unclear to what extent, and under which conditions, languages rely on each other{'}s data. To answer this question, we use TracIn (Pruthi et al., 2020), a training data attribution (TDA) method, to retrieve training samples from multilingual data that are most influential for test predictions in a given language. This allows us to analyse cross-lingual sharing mechanisms of MLMs from a new perspective. While previous work studied cross-lingual sharing at the model parameter level, we present the first approach to study it at the data level. We find that MLMs rely on data from multiple languages during fine-tuning and this reliance increases as fine-tuning progresses. We further find that training samples from other languages can both reinforce and complement the knowledge acquired from data of the test language itself. | [
"Choenni, Rochelle",
"Garrette, Dan",
"Shutova, Ekaterina"
] | How do languages influence each other? Studying cross-lingual data sharing during LM fine-tuning | emnlp-main.818 | 2305.13286 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.819.bib | https://aclanthology.org/2023.emnlp-main.819/ | @inproceedings{wang-etal-2023-coffee,
title = "{COFFEE}: Counterfactual Fairness for Personalized Text Generation in Explainable Recommendation",
author = "Wang, Nan and
Wang, Qifan and
Wang, Yi-Chia and
Sanjabi, Maziar and
Liu, Jingzhou and
Firooz, Hamed and
Wang, Hongning and
Nie, Shaoliang",
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.819",
doi = "10.18653/v1/2023.emnlp-main.819",
pages = "13258--13275",
abstract = "As language models become increasingly integrated into our digital lives, Personalized Text Generation (PTG) has emerged as a pivotal component with a wide range of applications. However, the bias inherent in user written text, often used for PTG model training, can inadvertently associate different levels of linguistic quality with users{'} protected attributes. The model can inherit the bias and perpetuate inequality in generating text w.r.t. users{'} protected attributes, leading to unfair treatment when serving users. In this work, we investigate fairness of PTG in the context of personalized explanation generation for recommendations. We first discuss the biases in generated explanations and their fairness implications. To promote fairness, we introduce a general framework to achieve measure-specific counterfactual fairness in explanation generation. Extensive experiments and human evaluations demonstrate the effectiveness of our method.",
}
| As language models become increasingly integrated into our digital lives, Personalized Text Generation (PTG) has emerged as a pivotal component with a wide range of applications. However, the bias inherent in user written text, often used for PTG model training, can inadvertently associate different levels of linguistic quality with users{'} protected attributes. The model can inherit the bias and perpetuate inequality in generating text w.r.t. users{'} protected attributes, leading to unfair treatment when serving users. In this work, we investigate fairness of PTG in the context of personalized explanation generation for recommendations. We first discuss the biases in generated explanations and their fairness implications. To promote fairness, we introduce a general framework to achieve measure-specific counterfactual fairness in explanation generation. Extensive experiments and human evaluations demonstrate the effectiveness of our method. | [
"Wang, Nan",
"Wang, Qifan",
"Wang, Yi-Chia",
"Sanjabi, Maziar",
"Liu, Jingzhou",
"Firooz, Hamed",
"Wang, Hongning",
"Nie, Shaoliang"
] | COFFEE: Counterfactual Fairness for Personalized Text Generation in Explainable Recommendation | emnlp-main.819 | 2210.15500 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.820.bib | https://aclanthology.org/2023.emnlp-main.820/ | @inproceedings{zhang-etal-2023-nameguess,
title = "{N}ame{G}uess: Column Name Expansion for Tabular Data",
author = "Zhang, Jiani and
Shen, Zhengyuan and
Srinivasan, Balasubramaniam and
Wang, Shen and
Rangwala, Huzefa and
Karypis, George",
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.820",
doi = "10.18653/v1/2023.emnlp-main.820",
pages = "13276--13290",
abstract = "Recent advances in large language models have revolutionized many sectors, including the database industry. One common challenge when dealing with large volumes of tabular data is the pervasive use of abbreviated column names, which can negatively impact performance on various data search, access, and understanding tasks. To address this issue, we introduce a new task, called NameGuess, to expand column names (used in database schema) as a natural language generation problem. We create a training dataset of 384K abbreviated-expanded column pairs using a new data fabrication method and a human-annotated evaluation benchmark that includes 9.2K examples from real-world tables. To tackle the complexities associated with polysemy and ambiguity in NameGuess, we enhance auto-regressive language models by conditioning on table content and column header names {--} yielding a fine-tuned model (with 2.7B parameters) that matches human performance. Furthermore, we conduct a comprehensive analysis (on multiple LLMs) to validate the effectiveness of table content in NameGuess and identify promising future opportunities. Code has been made available at https://github.com/amazon-science/nameguess.",
}
| Recent advances in large language models have revolutionized many sectors, including the database industry. One common challenge when dealing with large volumes of tabular data is the pervasive use of abbreviated column names, which can negatively impact performance on various data search, access, and understanding tasks. To address this issue, we introduce a new task, called NameGuess, to expand column names (used in database schema) as a natural language generation problem. We create a training dataset of 384K abbreviated-expanded column pairs using a new data fabrication method and a human-annotated evaluation benchmark that includes 9.2K examples from real-world tables. To tackle the complexities associated with polysemy and ambiguity in NameGuess, we enhance auto-regressive language models by conditioning on table content and column header names {--} yielding a fine-tuned model (with 2.7B parameters) that matches human performance. Furthermore, we conduct a comprehensive analysis (on multiple LLMs) to validate the effectiveness of table content in NameGuess and identify promising future opportunities. Code has been made available at https://github.com/amazon-science/nameguess. | [
"Zhang, Jiani",
"Shen, Zhengyuan",
"Srinivasan, Balasubramaniam",
"Wang, Shen",
"Rangwala, Huzefa",
"Karypis, George"
] | NameGuess: Column Name Expansion for Tabular Data | emnlp-main.820 | 2310.13196 | [
"https://github.com/amazon-science/nameguess"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.821.bib | https://aclanthology.org/2023.emnlp-main.821/ | @inproceedings{kew-etal-2023-bless,
title = "{BLESS}: Benchmarking Large Language Models on Sentence Simplification",
author = "Kew, Tannon and
Chi, Alison and
V{\'a}squez-Rodr{\'\i}guez, Laura and
Agrawal, Sweta and
Aumiller, Dennis and
Alva-Manchego, Fernando and
Shardlow, Matthew",
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.821",
doi = "10.18653/v1/2023.emnlp-main.821",
pages = "13291--13309",
abstract = "We present BLESS, a comprehensive performance benchmark of the most recent state-of-the-art Large Language Models (LLMs) on the task of text simplification (TS). We examine how well off-the-shelf LLMs can solve this challenging task, assessing a total of 44 models, differing in size, architecture, pre-training methods, and accessibility, on three test sets from different domains (Wikipedia, news, and medical) under a few-shot setting. Our analysis considers a suite of automatic metrics, as well as a large-scale quantitative investigation into the types of common edit operations performed by the different models. Furthermore, we perform a manual qualitative analysis on a subset of model outputs to better gauge the quality of the generated simplifications. Our evaluation indicates that the best LLMs, despite not being trained on TS perform comparably with state-of-the-art TS baselines. Additionally, we find that certain LLMs demonstrate a greater range and diversity of edit operations. Our performance benchmark will be available as a resource for the development of future TS methods and evaluation metrics.",
}
| We present BLESS, a comprehensive performance benchmark of the most recent state-of-the-art Large Language Models (LLMs) on the task of text simplification (TS). We examine how well off-the-shelf LLMs can solve this challenging task, assessing a total of 44 models, differing in size, architecture, pre-training methods, and accessibility, on three test sets from different domains (Wikipedia, news, and medical) under a few-shot setting. Our analysis considers a suite of automatic metrics, as well as a large-scale quantitative investigation into the types of common edit operations performed by the different models. Furthermore, we perform a manual qualitative analysis on a subset of model outputs to better gauge the quality of the generated simplifications. Our evaluation indicates that the best LLMs, despite not being trained on TS perform comparably with state-of-the-art TS baselines. Additionally, we find that certain LLMs demonstrate a greater range and diversity of edit operations. Our performance benchmark will be available as a resource for the development of future TS methods and evaluation metrics. | [
"Kew, Tannon",
"Chi, Alison",
"V{\\'a}squez-Rodr{\\'\\i}guez, Laura",
"Agrawal, Sweta",
"Aumiller, Dennis",
"Alva-Manchego, Fern",
"o",
"Shardlow, Matthew"
] | BLESS: Benchmarking Large Language Models on Sentence Simplification | emnlp-main.821 | 2310.15773 | [
"https://github.com/zurichnlp/bless"
] | https://huggingface.co/papers/2310.15773 | 0 | 0 | 0 | 7 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.822.bib | https://aclanthology.org/2023.emnlp-main.822/ | @inproceedings{gururaja-etal-2023-build,
title = "To Build Our Future, We Must Know Our Past: Contextualizing Paradigm Shifts in Natural Language Processing",
author = "Gururaja, Sireesh and
Bertsch, Amanda and
Na, Clara and
Widder, David and
Strubell, Emma",
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.822",
doi = "10.18653/v1/2023.emnlp-main.822",
pages = "13310--13325",
abstract = "NLP is in a period of disruptive change that is impacting our methodologies, funding sources, and public perception. In this work, we seek to understand how to shape our future by better understanding our past. We study factors that shape NLP as a field, including culture, incentives, and infrastructure by conducting long-form interviews with 26 NLP researchers of varying seniority, research area, institution, and social identity. Our interviewees identify cyclical patterns in the field, as well as new shifts without historical parallel, including changes in benchmark culture and software infrastructure. We complement this discussion with quantitative analysis of citation, authorship, and language use in the ACL Anthology over time. We conclude by discussing shared visions, concerns, and hopes for the future of NLP. We hope that this study of our field{'}s past and present can prompt informed discussion of our community{'}s implicit norms and more deliberate action to consciously shape the future.",
}
| NLP is in a period of disruptive change that is impacting our methodologies, funding sources, and public perception. In this work, we seek to understand how to shape our future by better understanding our past. We study factors that shape NLP as a field, including culture, incentives, and infrastructure by conducting long-form interviews with 26 NLP researchers of varying seniority, research area, institution, and social identity. Our interviewees identify cyclical patterns in the field, as well as new shifts without historical parallel, including changes in benchmark culture and software infrastructure. We complement this discussion with quantitative analysis of citation, authorship, and language use in the ACL Anthology over time. We conclude by discussing shared visions, concerns, and hopes for the future of NLP. We hope that this study of our field{'}s past and present can prompt informed discussion of our community{'}s implicit norms and more deliberate action to consciously shape the future. | [
"Gururaja, Sireesh",
"Bertsch, Am",
"a",
"Na, Clara",
"Widder, David",
"Strubell, Emma"
] | To Build Our Future, We Must Know Our Past: Contextualizing Paradigm Shifts in Natural Language Processing | emnlp-main.822 | 2310.07715 | [
""
] | https://huggingface.co/papers/2310.07715 | 0 | 0 | 0 | 5 | [] | [] | [] | 1 | Oral |
https://aclanthology.org/2023.emnlp-main.823.bib | https://aclanthology.org/2023.emnlp-main.823/ | @inproceedings{kanclerz-etal-2023-pals,
title = "{PALS}: Personalized Active Learning for Subjective Tasks in {NLP}",
author = "Kanclerz, Kamil and
Karanowski, Konrad and
Bielaniewicz, Julita and
Gruza, Marcin and
Mi{\l}kowski, Piotr and
Kocon, Jan and
Kazienko, Przemyslaw",
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.823",
doi = "10.18653/v1/2023.emnlp-main.823",
pages = "13326--13341",
abstract = "For subjective NLP problems, such as classification of hate speech, aggression, or emotions, personalized solutions can be exploited. Then, the learned models infer about the perception of the content independently for each reader. To acquire training data, texts are commonly randomly assigned to users for annotation, which is expensive and highly inefficient. Therefore, for the first time, we suggest applying an active learning paradigm in a personalized context to better learn individual preferences. It aims to alleviate the labeling effort by selecting more relevant training samples. In this paper, we present novel Personalized Active Learning techniques for Subjective NLP tasks (PALS) to either reduce the cost of the annotation process or to boost the learning effect. Our five new measures allow us to determine the relevance of a text in the context of learning users personal preferences. We validated them on three datasets: Wiki discussion texts individually labeled with aggression and toxicity, and on Unhealthy Conversations dataset. Our PALS techniques outperform random selection even by more than 30{\%}. They can also be used to reduce the number of necessary annotations while maintaining a given quality level. Personalized annotation assignments based on our controversy measure decrease the amount of data needed to just 25{\%}-40{\%} of the initial size.",
}
| For subjective NLP problems, such as classification of hate speech, aggression, or emotions, personalized solutions can be exploited. Then, the learned models infer about the perception of the content independently for each reader. To acquire training data, texts are commonly randomly assigned to users for annotation, which is expensive and highly inefficient. Therefore, for the first time, we suggest applying an active learning paradigm in a personalized context to better learn individual preferences. It aims to alleviate the labeling effort by selecting more relevant training samples. In this paper, we present novel Personalized Active Learning techniques for Subjective NLP tasks (PALS) to either reduce the cost of the annotation process or to boost the learning effect. Our five new measures allow us to determine the relevance of a text in the context of learning users personal preferences. We validated them on three datasets: Wiki discussion texts individually labeled with aggression and toxicity, and on Unhealthy Conversations dataset. Our PALS techniques outperform random selection even by more than 30{\%}. They can also be used to reduce the number of necessary annotations while maintaining a given quality level. Personalized annotation assignments based on our controversy measure decrease the amount of data needed to just 25{\%}-40{\%} of the initial size. | [
"Kanclerz, Kamil",
"Karanowski, Konrad",
"Bielaniewicz, Julita",
"Gruza, Marcin",
"Mi{\\l}kowski, Piotr",
"Kocon, Jan",
"Kazienko, Przemyslaw"
] | PALS: Personalized Active Learning for Subjective Tasks in NLP | emnlp-main.823 | null | [
"https://github.com/clarin-pl/personalized-nlp"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.824.bib | https://aclanthology.org/2023.emnlp-main.824/ | @inproceedings{chen-etal-2023-vistruct,
title = "{V}i{S}truct: Visual Structural Knowledge Extraction via Curriculum Guided Code-Vision Representation",
author = "Chen, Yangyi and
Wang, Xingyao and
Li, Manling and
Hoiem, Derek and
Ji, Heng",
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.824",
doi = "10.18653/v1/2023.emnlp-main.824",
pages = "13342--13357",
abstract = "State-of-the-art vision-language models (VLMs) still have limited performance in structural knowledge extraction, such as relations between objects. In this work, we present ViStruct, a training framework to learn VLMs for effective visual structural knowledge extraction. Two novel designs are incorporated. First, we propose to leverage the inherent structure of programming language to depict visual structural information. This approach enables explicit and consistent representation of visual structural information of multiple granularities, such as concepts, relations, and events, in a well-organized structured format. Second, we introduce curriculum-based learning for VLMs to progressively comprehend visual structures, from fundamental visual concepts to intricate event structures. Our intuition is that lower-level knowledge may contribute to complex visual structure understanding. Furthermore, we compile and release a collection of datasets tailored for visual structural knowledge extraction. We adopt a weakly-supervised approach to directly generate visual event structures from captions for ViStruct training, capitalizing on abundant image-caption pairs from the web. In experiments, we evaluate ViStruct on visual structure prediction tasks, demonstrating its effectiveness in improving the understanding of visual structures. The code will be made public to facilitate future research.",
}
| State-of-the-art vision-language models (VLMs) still have limited performance in structural knowledge extraction, such as relations between objects. In this work, we present ViStruct, a training framework to learn VLMs for effective visual structural knowledge extraction. Two novel designs are incorporated. First, we propose to leverage the inherent structure of programming language to depict visual structural information. This approach enables explicit and consistent representation of visual structural information of multiple granularities, such as concepts, relations, and events, in a well-organized structured format. Second, we introduce curriculum-based learning for VLMs to progressively comprehend visual structures, from fundamental visual concepts to intricate event structures. Our intuition is that lower-level knowledge may contribute to complex visual structure understanding. Furthermore, we compile and release a collection of datasets tailored for visual structural knowledge extraction. We adopt a weakly-supervised approach to directly generate visual event structures from captions for ViStruct training, capitalizing on abundant image-caption pairs from the web. In experiments, we evaluate ViStruct on visual structure prediction tasks, demonstrating its effectiveness in improving the understanding of visual structures. The code will be made public to facilitate future research. | [
"Chen, Yangyi",
"Wang, Xingyao",
"Li, Manling",
"Hoiem, Derek",
"Ji, Heng"
] | ViStruct: Visual Structural Knowledge Extraction via Curriculum Guided Code-Vision Representation | emnlp-main.824 | 2311.13258 | [
"https://github.com/yangyi-chen/vi-struct"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Oral |
|
https://aclanthology.org/2023.emnlp-main.825.bib | https://aclanthology.org/2023.emnlp-main.825/ | @inproceedings{jiang-etal-2023-llmlingua,
title = "{LLML}ingua: Compressing Prompts for Accelerated Inference of Large Language Models",
author = "Jiang, Huiqiang and
Wu, Qianhui and
Lin, Chin-Yew and
Yang, Yuqing and
Qiu, Lili",
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.825",
doi = "10.18653/v1/2023.emnlp-main.825",
pages = "13358--13376",
abstract = "Large language models (LLMs) have been applied in various applications due to their astonishing capabilities. With advancements in technologies such as chain-of-thought (CoT) prompting and in-context learning (ICL), the prompts fed to LLMs are becoming increasingly lengthy, even exceeding tens of thousands of tokens. To accelerate model inference and reduce cost, this paper presents LLMLingua, a coarse-to-fine prompt compression method that involves a budget controller to maintain semantic integrity under high compression ratios, a token-level iterative compression algorithm to better model the interdependence between compressed contents, and an instruction tuning based method for distribution alignment between language models. We conduct experiments and analysis over four datasets from different scenarios, i.e., GSM8K, BBH, ShareGPT, and Arxiv-March23; showing that the proposed approach yields state-of-the-art performance and allows for up to 20x compression with little performance loss.",
}
| Large language models (LLMs) have been applied in various applications due to their astonishing capabilities. With advancements in technologies such as chain-of-thought (CoT) prompting and in-context learning (ICL), the prompts fed to LLMs are becoming increasingly lengthy, even exceeding tens of thousands of tokens. To accelerate model inference and reduce cost, this paper presents LLMLingua, a coarse-to-fine prompt compression method that involves a budget controller to maintain semantic integrity under high compression ratios, a token-level iterative compression algorithm to better model the interdependence between compressed contents, and an instruction tuning based method for distribution alignment between language models. We conduct experiments and analysis over four datasets from different scenarios, i.e., GSM8K, BBH, ShareGPT, and Arxiv-March23; showing that the proposed approach yields state-of-the-art performance and allows for up to 20x compression with little performance loss. | [
"Jiang, Huiqiang",
"Wu, Qianhui",
"Lin, Chin-Yew",
"Yang, Yuqing",
"Qiu, Lili"
] | LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Models | emnlp-main.825 | 2310.05736 | [
"https://github.com/microsoft/LLMLingua"
] | https://huggingface.co/papers/2310.05736 | 2 | 4 | 0 | 5 | [] | [] | [
"microsoft/LLMLingua",
"Vincentt/LLMLingua"
] | 1 | Oral |
https://aclanthology.org/2023.emnlp-main.826.bib | https://aclanthology.org/2023.emnlp-main.826/ | @inproceedings{zhu-etal-2023-explain,
title = "{EXPLAIN}, {EDIT}, {GENERATE}: Rationale-Sensitive Counterfactual Data Augmentation for Multi-hop Fact Verification",
author = "Zhu, Yingjie and
Si, Jiasheng and
Zhao, Yibo and
Zhu, Haiyang and
Zhou, Deyu and
He, Yulan",
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.826",
doi = "10.18653/v1/2023.emnlp-main.826",
pages = "13377--13392",
abstract = "Automatic multi-hop fact verification task has gained significant attention in recent years. Despite impressive results, these well-designed models perform poorly on out-of-domain data. One possible solution is to augment the training data with counterfactuals, which are generated by minimally altering the causal features of the original data. However, current counterfactual data augmentation techniques fail to handle multi-hop fact verification due to their incapability to preserve the complex logical relationships within multiple correlated texts. In this paper, we overcome this limitation by developing a rationale-sensitive method to generate linguistically diverse and label-flipping counterfactuals while preserving logical relationships. In specific, the diverse and fluent counterfactuals are generated via an Explain-Edit-Generate architecture. Moreover, the checking and filtering modules are proposed to regularize the counterfactual data with logical relations and flipped labels. Experimental results show that the proposed approach outperforms the SOTA baselines and can generate linguistically diverse counterfactual data without disrupting their logical relationships.",
}
| Automatic multi-hop fact verification task has gained significant attention in recent years. Despite impressive results, these well-designed models perform poorly on out-of-domain data. One possible solution is to augment the training data with counterfactuals, which are generated by minimally altering the causal features of the original data. However, current counterfactual data augmentation techniques fail to handle multi-hop fact verification due to their incapability to preserve the complex logical relationships within multiple correlated texts. In this paper, we overcome this limitation by developing a rationale-sensitive method to generate linguistically diverse and label-flipping counterfactuals while preserving logical relationships. In specific, the diverse and fluent counterfactuals are generated via an Explain-Edit-Generate architecture. Moreover, the checking and filtering modules are proposed to regularize the counterfactual data with logical relations and flipped labels. Experimental results show that the proposed approach outperforms the SOTA baselines and can generate linguistically diverse counterfactual data without disrupting their logical relationships. | [
"Zhu, Yingjie",
"Si, Jiasheng",
"Zhao, Yibo",
"Zhu, Haiyang",
"Zhou, Deyu",
"He, Yulan"
] | EXPLAIN, EDIT, GENERATE: Rationale-Sensitive Counterfactual Data Augmentation for Multi-hop Fact Verification | emnlp-main.826 | 2310.14508 | [
"https://github.com/aaandy-zhu/race"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Oral |
|
https://aclanthology.org/2023.emnlp-main.827.bib | https://aclanthology.org/2023.emnlp-main.827/ | @inproceedings{opedal-etal-2023-exploration,
title = "An Exploration of Left-Corner Transformations",
author = "Opedal, Andreas and
Tsipidi, Eleftheria and
Pimentel, Tiago and
Cotterell, Ryan and
Vieira, Tim",
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.827",
doi = "10.18653/v1/2023.emnlp-main.827",
pages = "13393--13427",
abstract = "The left-corner transformation (Rosenkrantz and Lewis, 1970) is used to remove left recursion from context-free grammars, which is an important step towards making the grammar parsable top-down with simple techniques. This paper generalizes prior left-corner transformations to support semiring-weighted production rules and to provide finer-grained control over which left corners may be moved. Our generalized left-corner transformation (GLCT) arose from unifying the left-corner transformation and speculation transformation (Eisner and Blatz, 2007), originally for logic programming. Our new transformation and speculation define equivalent weighted languages. Yet, their derivation trees are structurally different in an important way: GLCT replaces left recursion with right recursion, and speculation does not. We also provide several technical results regarding the formal relationships between the outputs of GLCT, speculation, and the original grammar. Lastly, we empirically investigate the efficiency of GLCT for left-recursion elimination from grammars of nine languages. Code: https://github.com/rycolab/left-corner",
}
| The left-corner transformation (Rosenkrantz and Lewis, 1970) is used to remove left recursion from context-free grammars, which is an important step towards making the grammar parsable top-down with simple techniques. This paper generalizes prior left-corner transformations to support semiring-weighted production rules and to provide finer-grained control over which left corners may be moved. Our generalized left-corner transformation (GLCT) arose from unifying the left-corner transformation and speculation transformation (Eisner and Blatz, 2007), originally for logic programming. Our new transformation and speculation define equivalent weighted languages. Yet, their derivation trees are structurally different in an important way: GLCT replaces left recursion with right recursion, and speculation does not. We also provide several technical results regarding the formal relationships between the outputs of GLCT, speculation, and the original grammar. Lastly, we empirically investigate the efficiency of GLCT for left-recursion elimination from grammars of nine languages. Code: https://github.com/rycolab/left-corner | [
"Opedal, Andreas",
"Tsipidi, Eleftheria",
"Pimentel, Tiago",
"Cotterell, Ryan",
"Vieira, Tim"
] | An Exploration of Left-Corner Transformations | emnlp-main.827 | 2311.16258 | [
""
] | https://huggingface.co/papers/2311.16258 | 1 | 1 | 0 | 5 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.828.bib | https://aclanthology.org/2023.emnlp-main.828/ | @inproceedings{wu-etal-2023-characterizing,
title = "Characterizing and Verifying Scientific Claims: Qualitative Causal Structure is All You Need",
author = "Wu, Jinxuan and
Chao, Wenhan and
Zhou, Xian and
Luo, Zhunchen",
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.828",
doi = "10.18653/v1/2023.emnlp-main.828",
pages = "13428--13439",
abstract = "A scientific claim typically begins with the formulation of a research question or hypothesis, which is a tentative statement or proposition about a phenomenon or relationship between variables. Within the realm of scientific claim verification, considerable research efforts have been dedicated to attention architectures and leveraging the text comprehension capabilities of Pre-trained Language Models (PLMs), yielding promising performances. However, these models overlook the causal structure information inherent in scientific claims, thereby failing to establish a comprehensive chain of causal inference. This paper delves into the exploration to highlight the crucial role of qualitative causal structure in characterizing and verifying scientific claims based on evidence. We organize the qualitative causal structure into a heterogeneous graph and propose a novel attention-based graph neural network model to facilitate causal reasoning across relevant causally-potent factors. Our experiments demonstrate that by solely utilizing the qualitative causal structure, the proposed model achieves comparable performance to PLM-based models. Furthermore, by incorporating semantic features, our model outperforms state-of-the-art approaches comprehensively.",
}
| A scientific claim typically begins with the formulation of a research question or hypothesis, which is a tentative statement or proposition about a phenomenon or relationship between variables. Within the realm of scientific claim verification, considerable research efforts have been dedicated to attention architectures and leveraging the text comprehension capabilities of Pre-trained Language Models (PLMs), yielding promising performances. However, these models overlook the causal structure information inherent in scientific claims, thereby failing to establish a comprehensive chain of causal inference. This paper delves into the exploration to highlight the crucial role of qualitative causal structure in characterizing and verifying scientific claims based on evidence. We organize the qualitative causal structure into a heterogeneous graph and propose a novel attention-based graph neural network model to facilitate causal reasoning across relevant causally-potent factors. Our experiments demonstrate that by solely utilizing the qualitative causal structure, the proposed model achieves comparable performance to PLM-based models. Furthermore, by incorporating semantic features, our model outperforms state-of-the-art approaches comprehensively. | [
"Wu, Jinxuan",
"Chao, Wenhan",
"Zhou, Xian",
"Luo, Zhunchen"
] | Characterizing and Verifying Scientific Claims: Qualitative Causal Structure is All You Need | emnlp-main.828 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.829.bib | https://aclanthology.org/2023.emnlp-main.829/ | @inproceedings{dobler-de-melo-2023-focus,
title = "{FOCUS}: Effective Embedding Initialization for Monolingual Specialization of Multilingual Models",
author = "Dobler, Konstantin and
de Melo, Gerard",
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.829",
doi = "10.18653/v1/2023.emnlp-main.829",
pages = "13440--13454",
abstract = "Using model weights pretrained on a high-resource language as a warm start can reduce the need for data and compute to obtain high-quality language models for other, especially low-resource, languages. However, if we want to use a new tokenizer specialized for the target language, we cannot transfer the source model{'}s embedding matrix. In this paper, we propose FOCUS - **F**ast **O**verlapping Token **C**ombinations **U**sing **S**parsemax, a novel embedding initialization method that effectively initializes the embedding matrix for a new tokenizer based on information in the source model{'}s embedding matrix. FOCUS represents newly added tokens as combinations of tokens in the overlap of the source and target vocabularies. The overlapping tokens are selected based on semantic similarity in an auxiliary static token embedding space. We focus our study on using the multilingual XLM-R as a source model and empirically show that FOCUS outperforms random initialization and previous work on language modeling and on a range of downstream tasks (NLI, QA, and NER). We publish our model checkpoints and code on GitHub.",
}
| Using model weights pretrained on a high-resource language as a warm start can reduce the need for data and compute to obtain high-quality language models for other, especially low-resource, languages. However, if we want to use a new tokenizer specialized for the target language, we cannot transfer the source model{'}s embedding matrix. In this paper, we propose FOCUS - **F**ast **O**verlapping Token **C**ombinations **U**sing **S**parsemax, a novel embedding initialization method that effectively initializes the embedding matrix for a new tokenizer based on information in the source model{'}s embedding matrix. FOCUS represents newly added tokens as combinations of tokens in the overlap of the source and target vocabularies. The overlapping tokens are selected based on semantic similarity in an auxiliary static token embedding space. We focus our study on using the multilingual XLM-R as a source model and empirically show that FOCUS outperforms random initialization and previous work on language modeling and on a range of downstream tasks (NLI, QA, and NER). We publish our model checkpoints and code on GitHub. | [
"Dobler, Konstantin",
"de Melo, Gerard"
] | FOCUS: Effective Embedding Initialization for Monolingual Specialization of Multilingual Models | emnlp-main.829 | 2305.14481 | [
"https://github.com/konstantinjdobler/focus"
] | https://huggingface.co/papers/2305.14481 | 1 | 1 | 0 | 2 | [
"konstantindobler/xlm-roberta-base-focus-german",
"konstantindobler/xlm-roberta-base-focus-hausa",
"konstantindobler/xlm-roberta-base-focus-isixhosa",
"konstantindobler/xlm-roberta-base-focus-arabic",
"konstantindobler/xlm-roberta-base-focus-extend-isixhosa",
"konstantindobler/xlm-roberta-base-focus-kiswahili",
"konstantindobler/xlm-roberta-base-focus-extend-german",
"konstantindobler/xlm-roberta-base-focus-extend-kiswahili",
"konstantindobler/xlm-roberta-base-focus-extend-arabic",
"konstantindobler/xlm-roberta-base-focus-extend-hausa"
] | [] | [] | 1 | Oral |
https://aclanthology.org/2023.emnlp-main.830.bib | https://aclanthology.org/2023.emnlp-main.830/ | @inproceedings{wang-etal-2023-bytesized32,
title = "{B}yte{S}ized32: A Corpus and Challenge Task for Generating Task-Specific World Models Expressed as Text Games",
author = "Wang, Ruoyao and
Todd, Graham and
Yuan, Xingdi and
Xiao, Ziang and
C{\^o}t{\'e}, Marc-Alexandre and
Jansen, Peter",
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.830",
doi = "10.18653/v1/2023.emnlp-main.830",
pages = "13455--13471",
abstract = "In this work we investigate the capacity of language models to generate explicit, interpretable, and interactive world models of scientific and common-sense reasoning tasks. We operationalize this as a task of generating text games, expressed as hundreds of lines of Python code. To facilitate this task, we introduce ByteSized32, a corpus of 32 reasoning-focused text games totalling 20k lines of Python code. We empirically demonstrate that GPT-4 can use these games as templates for single-shot in-context learning, successfully producing runnable games on unseen topics in 28{\%} of cases. When allowed to self-reflect on program errors, game runnability substantially increases to 58{\%}. While evaluating simulation fidelity is labor intensive, we introduce a suite of automated metrics to assess game fidelity, technical validity, adherence to task specifications, and winnability, showing a high-degree of agreement with expert human ratings. We pose this as a challenge task to spur further development at the juncture of world modeling and code generation.",
}
| In this work we investigate the capacity of language models to generate explicit, interpretable, and interactive world models of scientific and common-sense reasoning tasks. We operationalize this as a task of generating text games, expressed as hundreds of lines of Python code. To facilitate this task, we introduce ByteSized32, a corpus of 32 reasoning-focused text games totalling 20k lines of Python code. We empirically demonstrate that GPT-4 can use these games as templates for single-shot in-context learning, successfully producing runnable games on unseen topics in 28{\%} of cases. When allowed to self-reflect on program errors, game runnability substantially increases to 58{\%}. While evaluating simulation fidelity is labor intensive, we introduce a suite of automated metrics to assess game fidelity, technical validity, adherence to task specifications, and winnability, showing a high-degree of agreement with expert human ratings. We pose this as a challenge task to spur further development at the juncture of world modeling and code generation. | [
"Wang, Ruoyao",
"Todd, Graham",
"Yuan, Xingdi",
"Xiao, Ziang",
"C{\\^o}t{\\'e}, Marc-Alex",
"re",
"Jansen, Peter"
] | ByteSized32: A Corpus and Challenge Task for Generating Task-Specific World Models Expressed as Text Games | emnlp-main.830 | 2305.14879 | [
"https://github.com/cognitiveailab/bytesized32"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.831.bib | https://aclanthology.org/2023.emnlp-main.831/ | @inproceedings{an-etal-2023-skill,
title = "Skill-Based Few-Shot Selection for In-Context Learning",
author = "An, Shengnan and
Zhou, Bo and
Lin, Zeqi and
Fu, Qiang and
Chen, Bei and
Zheng, Nanning and
Chen, Weizhu and
Lou, Jian-Guang",
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.831",
doi = "10.18653/v1/2023.emnlp-main.831",
pages = "13472--13492",
abstract = "*In-context learning* is the paradigm that adapts large language models to downstream tasks by providing a few examples. *Few-shot selection*{---}selecting appropriate examples for each test instance separately{---}is important for in-context learning. In this paper, we propose **Skill-KNN**, a skill-based few-shot selection method for in-context learning. The key advantages of Skill-KNN include: (1) it addresses the problem that existing methods based on pre-trained embeddings can be easily biased by surface natural language features that are not important for the target task; (2) it does not require training or fine-tuning of any models, making it suitable for frequently expanding or changing example banks. The key insight is to optimize the inputs fed into the embedding model, rather than tuning the model itself. Technically, Skill-KNN generates the skill-based descriptions for each test case and candidate example by utilizing a pre-processing few-shot prompting, thus eliminating unimportant surface features. Experimental results across five cross-domain semantic parsing datasets and six backbone models show that Skill-KNN significantly outperforms existing methods.",
}
| *In-context learning* is the paradigm that adapts large language models to downstream tasks by providing a few examples. *Few-shot selection*{---}selecting appropriate examples for each test instance separately{---}is important for in-context learning. In this paper, we propose **Skill-KNN**, a skill-based few-shot selection method for in-context learning. The key advantages of Skill-KNN include: (1) it addresses the problem that existing methods based on pre-trained embeddings can be easily biased by surface natural language features that are not important for the target task; (2) it does not require training or fine-tuning of any models, making it suitable for frequently expanding or changing example banks. The key insight is to optimize the inputs fed into the embedding model, rather than tuning the model itself. Technically, Skill-KNN generates the skill-based descriptions for each test case and candidate example by utilizing a pre-processing few-shot prompting, thus eliminating unimportant surface features. Experimental results across five cross-domain semantic parsing datasets and six backbone models show that Skill-KNN significantly outperforms existing methods. | [
"An, Shengnan",
"Zhou, Bo",
"Lin, Zeqi",
"Fu, Qiang",
"Chen, Bei",
"Zheng, Nanning",
"Chen, Weizhu",
"Lou, Jian-Guang"
] | Skill-Based Few-Shot Selection for In-Context Learning | emnlp-main.831 | 2305.14210 | [
""
] | https://huggingface.co/papers/2305.14210 | 2 | 0 | 0 | 8 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.832.bib | https://aclanthology.org/2023.emnlp-main.832/ | @inproceedings{menon-etal-2023-mantle,
title = "{M}a{N}t{LE}: Model-agnostic Natural Language Explainer",
author = "Menon, Rakesh and
Zaman, Kerem and
Srivastava, Shashank",
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.832",
doi = "10.18653/v1/2023.emnlp-main.832",
pages = "13493--13511",
abstract = "Understanding the internal reasoning behind the predictions of machine learning systems is increasingly vital, given their rising adoption and acceptance. While previous approaches, such as LIME generate algorithmic explanations by attributing importance to input features for individual examples, recent research indicates that practitioners prefer examining language explanations that explain sub-groups of examples (Lakkaraju et al., 2022). In this paper, we introduce MaNtLE, a model-agnostic natural language explainer that analyzes a set of classifier predictions and generates faithful natural language explanations of classifier rationale for structured classification tasks. MaNtLE uses multi-task training on thousands of synthetic classification tasks to generate faithful explanations. Our experiments indicate that, on average, MaNtLE-generated explanations are at least 11{\%} more faithful compared to LIME and Anchors explanations across three tasks. Human evaluations demonstrate that users can better predict model behavior using explanations from MaNtLE compared to other techniques.",
}
| Understanding the internal reasoning behind the predictions of machine learning systems is increasingly vital, given their rising adoption and acceptance. While previous approaches, such as LIME generate algorithmic explanations by attributing importance to input features for individual examples, recent research indicates that practitioners prefer examining language explanations that explain sub-groups of examples (Lakkaraju et al., 2022). In this paper, we introduce MaNtLE, a model-agnostic natural language explainer that analyzes a set of classifier predictions and generates faithful natural language explanations of classifier rationale for structured classification tasks. MaNtLE uses multi-task training on thousands of synthetic classification tasks to generate faithful explanations. Our experiments indicate that, on average, MaNtLE-generated explanations are at least 11{\%} more faithful compared to LIME and Anchors explanations across three tasks. Human evaluations demonstrate that users can better predict model behavior using explanations from MaNtLE compared to other techniques. | [
"Menon, Rakesh",
"Zaman, Kerem",
"Srivastava, Shashank"
] | MaNtLE: Model-agnostic Natural Language Explainer | emnlp-main.832 | 2305.12995 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.833.bib | https://aclanthology.org/2023.emnlp-main.833/ | @inproceedings{chen-etal-2023-ptp,
title = "{PTP}: Boosting Stability and Performance of Prompt Tuning with Perturbation-Based Regularizer",
author = "Chen, Lichang and
Chen, Jiuhai and
Huang, Heng and
Cheng, Minhao",
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.833",
doi = "10.18653/v1/2023.emnlp-main.833",
pages = "13512--13525",
abstract = "Recent studies show that prompt tuning can better leverage the power of large language models than fine-tuning on downstream natural language understanding tasks. However, the existing prompt tuning methods have training instability issues, as the variance of scores under different random seeds is quite large. To address this critical problem, we first investigate and find that the loss landscape of vanilla prompt tuning is precipitous when it is visualized, where a slight change of input data can cause a big fluctuation in the loss landscape. This is an essential factor that leads to the instability of prompt tuning. Based on this observation, we introduce perturbation-based regularizers, which can smooth the loss landscape, into prompt tuning. We propose a new algorithm, called Prompt Tuning with Perturbation-based regularizer (PTP), which can not only alleviate training instability dramatically but also boost the performance of prompt tuning. We design two kinds of perturbation-based regularizers, including random-noise-based and adversarial-based. In particular, our proposed perturbations are flexible on both text space and embedding space. Extensive experiments show the effectiveness of our proposed methods in stabilizing the training. Our new algorithms improve the state-of-the-art prompt tuning methods by 1.94{\%} and 2.34{\%} on SuperGLUE and FewGLUE benchmarks, respectively.",
}
| Recent studies show that prompt tuning can better leverage the power of large language models than fine-tuning on downstream natural language understanding tasks. However, the existing prompt tuning methods have training instability issues, as the variance of scores under different random seeds is quite large. To address this critical problem, we first investigate and find that the loss landscape of vanilla prompt tuning is precipitous when it is visualized, where a slight change of input data can cause a big fluctuation in the loss landscape. This is an essential factor that leads to the instability of prompt tuning. Based on this observation, we introduce perturbation-based regularizers, which can smooth the loss landscape, into prompt tuning. We propose a new algorithm, called Prompt Tuning with Perturbation-based regularizer (PTP), which can not only alleviate training instability dramatically but also boost the performance of prompt tuning. We design two kinds of perturbation-based regularizers, including random-noise-based and adversarial-based. In particular, our proposed perturbations are flexible on both text space and embedding space. Extensive experiments show the effectiveness of our proposed methods in stabilizing the training. Our new algorithms improve the state-of-the-art prompt tuning methods by 1.94{\%} and 2.34{\%} on SuperGLUE and FewGLUE benchmarks, respectively. | [
"Chen, Lichang",
"Chen, Jiuhai",
"Huang, Heng",
"Cheng, Minhao"
] | PTP: Boosting Stability and Performance of Prompt Tuning with Perturbation-Based Regularizer | emnlp-main.833 | 2305.02423 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.834.bib | https://aclanthology.org/2023.emnlp-main.834/ | @inproceedings{elgaar-amiri-2023-ling,
title = "Ling-{CL}: Understanding {NLP} Models through Linguistic Curricula",
author = "Elgaar, Mohamed and
Amiri, Hadi",
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.834",
doi = "10.18653/v1/2023.emnlp-main.834",
pages = "13526--13542",
abstract = "We employ a characterization of linguistic complexity from psycholinguistic and language acquisition research to develop data-driven curricula to understand the underlying linguistic knowledge that models learn to address NLP tasks. The novelty of our approach is in the development of linguistic curricula derived from data, existing knowledge about linguistic complexity, and model behavior during training. Through the evaluation of several benchmark NLP datasets, our curriculum learning approaches identify sets of linguistic metrics (indices) that inform the challenges and reasoning required to address each task. Our work will inform future research in all NLP areas, allowing linguistic complexity to be considered early in the research and development process. In addition, our work prompts an examination of gold standards and fair evaluation in NLP.",
}
| We employ a characterization of linguistic complexity from psycholinguistic and language acquisition research to develop data-driven curricula to understand the underlying linguistic knowledge that models learn to address NLP tasks. The novelty of our approach is in the development of linguistic curricula derived from data, existing knowledge about linguistic complexity, and model behavior during training. Through the evaluation of several benchmark NLP datasets, our curriculum learning approaches identify sets of linguistic metrics (indices) that inform the challenges and reasoning required to address each task. Our work will inform future research in all NLP areas, allowing linguistic complexity to be considered early in the research and development process. In addition, our work prompts an examination of gold standards and fair evaluation in NLP. | [
"Elgaar, Mohamed",
"Amiri, Hadi"
] | Ling-CL: Understanding NLP Models through Linguistic Curricula | emnlp-main.834 | 2310.20121 | [
"https://github.com/clu-uml/ling-cl"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.835.bib | https://aclanthology.org/2023.emnlp-main.835/ | @inproceedings{vamvas-sennrich-2023-towards,
title = "Towards Unsupervised Recognition of Token-level Semantic Differences in Related Documents",
author = "Vamvas, Jannis and
Sennrich, Rico",
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.835",
doi = "10.18653/v1/2023.emnlp-main.835",
pages = "13543--13552",
abstract = "Automatically highlighting words that cause semantic differences between two documents could be useful for a wide range of applications. We formulate recognizing semantic differences (RSD) as a token-level regression task and study three unsupervised approaches that rely on a masked language model. To assess the approaches, we begin with basic English sentences and gradually move to more complex, cross-lingual document pairs. Our results show that an approach based on word alignment and sentence-level contrastive learning has a robust correlation to gold labels. However, all unsupervised approaches still leave a large margin of improvement.",
}
| Automatically highlighting words that cause semantic differences between two documents could be useful for a wide range of applications. We formulate recognizing semantic differences (RSD) as a token-level regression task and study three unsupervised approaches that rely on a masked language model. To assess the approaches, we begin with basic English sentences and gradually move to more complex, cross-lingual document pairs. Our results show that an approach based on word alignment and sentence-level contrastive learning has a robust correlation to gold labels. However, all unsupervised approaches still leave a large margin of improvement. | [
"Vamvas, Jannis",
"Sennrich, Rico"
] | Towards Unsupervised Recognition of Token-level Semantic Differences in Related Documents | emnlp-main.835 | 2305.13303 | [
"https://github.com/zurichnlp/recognizing-semantic-differences"
] | https://huggingface.co/papers/2305.13303 | 1 | 0 | 0 | 2 | [
"ZurichNLP/unsup-simcse-xlm-roberta-base"
] | [
"ZurichNLP/rsd-ists-2016"
] | [
"ZurichNLP/unsupervised-semantic-diff"
] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.836.bib | https://aclanthology.org/2023.emnlp-main.836/ | @inproceedings{tan-monz-2023-towards,
title = "Towards a Better Understanding of Variations in Zero-Shot Neural Machine Translation Performance",
author = "Tan, Shaomu 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.836",
doi = "10.18653/v1/2023.emnlp-main.836",
pages = "13553--13568",
abstract = "Multilingual Neural Machine Translation (MNMT) facilitates knowledge sharing but often suffers from poor zero-shot (ZS) translation qualities. While prior work has explored the causes of overall low zero-shot translation qualities, our work introduces a fresh perspective: the presence of significant variations in zero-shot performance. This suggests that MNMT does not uniformly exhibit poor zero-shot capability; instead, certain translation directions yield reasonable results. Through systematic experimentation, spanning 1,560 language directions across 40 languages, we identify three key factors contributing to high variations in ZS NMT performance: 1) target-side translation quality, 2) vocabulary overlap, and 3) linguistic properties. Our findings highlight that the target side translation quality is the most influential factor, with vocabulary overlap consistently impacting zero-shot capabilities. Additionally, linguistic properties, such as language family and writing system, play a role, particularly with smaller models. Furthermore, we suggest that the off-target issue is a symptom of inadequate performance, emphasizing that zero-shot translation challenges extend beyond addressing the off-target problem. To support future research, we release the data and models as a benchmark for the study of ZS NMT.",
}
| Multilingual Neural Machine Translation (MNMT) facilitates knowledge sharing but often suffers from poor zero-shot (ZS) translation qualities. While prior work has explored the causes of overall low zero-shot translation qualities, our work introduces a fresh perspective: the presence of significant variations in zero-shot performance. This suggests that MNMT does not uniformly exhibit poor zero-shot capability; instead, certain translation directions yield reasonable results. Through systematic experimentation, spanning 1,560 language directions across 40 languages, we identify three key factors contributing to high variations in ZS NMT performance: 1) target-side translation quality, 2) vocabulary overlap, and 3) linguistic properties. Our findings highlight that the target side translation quality is the most influential factor, with vocabulary overlap consistently impacting zero-shot capabilities. Additionally, linguistic properties, such as language family and writing system, play a role, particularly with smaller models. Furthermore, we suggest that the off-target issue is a symptom of inadequate performance, emphasizing that zero-shot translation challenges extend beyond addressing the off-target problem. To support future research, we release the data and models as a benchmark for the study of ZS NMT. | [
"Tan, Shaomu",
"Monz, Christof"
] | Towards a Better Understanding of Variations in Zero-Shot Neural Machine Translation Performance | emnlp-main.836 | 2310.10385 | [
"https://github.com/Smu-Tan/ZS-NMT-Variations"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.837.bib | https://aclanthology.org/2023.emnlp-main.837/ | @inproceedings{tonglet-etal-2023-seer,
title = "{SEER} : A Knapsack approach to Exemplar Selection for In-Context {H}ybrid{QA}",
author = "Tonglet, Jonathan and
Reusens, Manon and
Borchert, Philipp and
Baesens, Bart",
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.837",
doi = "10.18653/v1/2023.emnlp-main.837",
pages = "13569--13583",
abstract = "Question answering over hybrid contexts is a complex task, which requires the combination of information extracted from unstructured texts and structured tables in various ways. Recently, In-Context Learning demonstrated significant performance advances for reasoning tasks. In this paradigm, a large language model performs predictions based on a small set of supporting exemplars. The performance of In-Context Learning depends heavily on the selection procedure of the supporting exemplars, particularly in the case of HybridQA, where considering the diversity of reasoning chains and the large size of the hybrid contexts becomes crucial. In this work, we present Selection of ExEmplars for hybrid Reasoning (SEER), a novel method for selecting a set of exemplars that is both representative and diverse. The key novelty of SEER is that it formulates exemplar selection as a Knapsack Integer Linear Program. The Knapsack framework provides the flexibility to incorporate diversity constraints that prioritize exemplars with desirable attributes, and capacity constraints that ensure that the prompt size respects the provided capacity budgets. The effectiveness of SEER is demonstrated on FinQA and TAT-QA, two real-world benchmarks for HybridQA, where it outperforms previous exemplar selection methods.",
}
| Question answering over hybrid contexts is a complex task, which requires the combination of information extracted from unstructured texts and structured tables in various ways. Recently, In-Context Learning demonstrated significant performance advances for reasoning tasks. In this paradigm, a large language model performs predictions based on a small set of supporting exemplars. The performance of In-Context Learning depends heavily on the selection procedure of the supporting exemplars, particularly in the case of HybridQA, where considering the diversity of reasoning chains and the large size of the hybrid contexts becomes crucial. In this work, we present Selection of ExEmplars for hybrid Reasoning (SEER), a novel method for selecting a set of exemplars that is both representative and diverse. The key novelty of SEER is that it formulates exemplar selection as a Knapsack Integer Linear Program. The Knapsack framework provides the flexibility to incorporate diversity constraints that prioritize exemplars with desirable attributes, and capacity constraints that ensure that the prompt size respects the provided capacity budgets. The effectiveness of SEER is demonstrated on FinQA and TAT-QA, two real-world benchmarks for HybridQA, where it outperforms previous exemplar selection methods. | [
"Tonglet, Jonathan",
"Reusens, Manon",
"Borchert, Philipp",
"Baesens, Bart"
] | SEER : A Knapsack approach to Exemplar Selection for In-Context HybridQA | emnlp-main.837 | 2310.06675 | [
"https://github.com/jtonglet/seer"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.838.bib | https://aclanthology.org/2023.emnlp-main.838/ | @inproceedings{jang-etal-2023-conversation,
title = "Conversation Chronicles: Towards Diverse Temporal and Relational Dynamics in Multi-Session Conversations",
author = "Jang, Jihyoung and
Boo, Minseong and
Kim, Hyounghun",
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.838",
doi = "10.18653/v1/2023.emnlp-main.838",
pages = "13584--13606",
abstract = "In the field of natural language processing, open-domain chatbots have emerged as an important research topic. However, a major limitation of existing open-domain chatbot research is its singular focus on short single-session dialogue, neglecting the potential need for understanding contextual information in multiple consecutive sessions that precede an ongoing dialogue. Among the elements that compose the context in multi-session conversation settings, the time intervals between sessions and the relationships between speakers would be particularly important. Despite their importance, current research efforts have not sufficiently addressed these dialogical components. In this paper, we introduce a new 1M multi-session dialogue dataset, called Conversation Chronicles, for implementing a long-term conversation setup in which time intervals and fine-grained speaker relationships are incorporated. Following recent works, we exploit a large language model to produce the data. The extensive human evaluation shows that dialogue episodes in Conversation Chronicles reflect those properties while maintaining coherent and consistent interactions across all the sessions. We also propose a dialogue model, called ReBot, which consists of chronological summarization and dialogue generation modules using only around 630M parameters. When trained on Conversation Chronicles, ReBot demonstrates long-term context understanding with a high human engagement score.",
}
| In the field of natural language processing, open-domain chatbots have emerged as an important research topic. However, a major limitation of existing open-domain chatbot research is its singular focus on short single-session dialogue, neglecting the potential need for understanding contextual information in multiple consecutive sessions that precede an ongoing dialogue. Among the elements that compose the context in multi-session conversation settings, the time intervals between sessions and the relationships between speakers would be particularly important. Despite their importance, current research efforts have not sufficiently addressed these dialogical components. In this paper, we introduce a new 1M multi-session dialogue dataset, called Conversation Chronicles, for implementing a long-term conversation setup in which time intervals and fine-grained speaker relationships are incorporated. Following recent works, we exploit a large language model to produce the data. The extensive human evaluation shows that dialogue episodes in Conversation Chronicles reflect those properties while maintaining coherent and consistent interactions across all the sessions. We also propose a dialogue model, called ReBot, which consists of chronological summarization and dialogue generation modules using only around 630M parameters. When trained on Conversation Chronicles, ReBot demonstrates long-term context understanding with a high human engagement score. | [
"Jang, Jihyoung",
"Boo, Minseong",
"Kim, Hyounghun"
] | Conversation Chronicles: Towards Diverse Temporal and Relational Dynamics in Multi-Session Conversations | emnlp-main.838 | 2310.13420 | [
""
] | https://huggingface.co/papers/2310.13420 | 1 | 0 | 0 | 3 | [
"jihyoung/rebot-generation",
"jihyoung/rebot-summarization"
] | [] | [] | 1 | Oral |
https://aclanthology.org/2023.emnlp-main.839.bib | https://aclanthology.org/2023.emnlp-main.839/ | @inproceedings{hasegawa-etal-2023-duet,
title = "{D}ue{T}: Image-Text Contrastive Transfer Learning with Dual-adapter Tuning",
author = "Hasegawa, Taku and
Nishida, Kyosuke and
Maeda, Koki and
Saito, Kuniko",
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.839",
doi = "10.18653/v1/2023.emnlp-main.839",
pages = "13607--13624",
abstract = "This paper presents DueT, a novel transfer learning method for vision and language models built by contrastive learning. In DueT, adapters are inserted into the image and text encoders, which have been initialized using models pre-trained on uni-modal corpora and then frozen. By training only these adapters, DueT enables efficient learning with a reduced number of trainable parameters. Moreover, unlike traditional adapters, those in DueT are equipped with a gating mechanism, enabling effective transfer and connection of knowledge acquired from pre-trained uni-modal encoders while preventing catastrophic forgetting. We report that DueT outperformed simple fine-tuning, the conventional method fixing only the image encoder and training only the text encoder, and the LoRA-based adapter method in accuracy and parameter efficiency for 0-shot image and text retrieval in both English and Japanese domains.",
}
| This paper presents DueT, a novel transfer learning method for vision and language models built by contrastive learning. In DueT, adapters are inserted into the image and text encoders, which have been initialized using models pre-trained on uni-modal corpora and then frozen. By training only these adapters, DueT enables efficient learning with a reduced number of trainable parameters. Moreover, unlike traditional adapters, those in DueT are equipped with a gating mechanism, enabling effective transfer and connection of knowledge acquired from pre-trained uni-modal encoders while preventing catastrophic forgetting. We report that DueT outperformed simple fine-tuning, the conventional method fixing only the image encoder and training only the text encoder, and the LoRA-based adapter method in accuracy and parameter efficiency for 0-shot image and text retrieval in both English and Japanese domains. | [
"Hasegawa, Taku",
"Nishida, Kyosuke",
"Maeda, Koki",
"Saito, Kuniko"
] | DueT: Image-Text Contrastive Transfer Learning with Dual-adapter Tuning | emnlp-main.839 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.840.bib | https://aclanthology.org/2023.emnlp-main.840/ | @inproceedings{jung-etal-2023-towards,
title = "Towards a Unified Conversational Recommendation System: Multi-task Learning via Contextualized Knowledge Distillation",
author = "Jung, Yeongseo and
Jung, Eunseo and
Chen, 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.840",
doi = "10.18653/v1/2023.emnlp-main.840",
pages = "13625--13637",
abstract = "In Conversational Recommendation System (CRS), an agent is asked to recommend a set of items to users within natural language conversations. To address the need for both conversational capability and personalized recommendations, prior works have utilized separate recommendation and dialogue modules. However, such approach inevitably results in a discrepancy between recommendation results and generated responses. To bridge the gap, we propose a multi-task learning for a unified CRS, where a single model jointly learns both tasks via Contextualized Knowledge Distillation (ConKD). We introduce two versions of ConKD: hard gate and soft gate. The former selectively gates between two task-specific teachers, while the latter integrates knowledge from both teachers. Our gates are computed on-the-fly in a context-specific manner, facilitating flexible integration of relevant knowledge. Extensive experiments demonstrate that our single model significantly improves recommendation performance while enhancing fluency, and achieves comparable results in terms of diversity.",
}
| In Conversational Recommendation System (CRS), an agent is asked to recommend a set of items to users within natural language conversations. To address the need for both conversational capability and personalized recommendations, prior works have utilized separate recommendation and dialogue modules. However, such approach inevitably results in a discrepancy between recommendation results and generated responses. To bridge the gap, we propose a multi-task learning for a unified CRS, where a single model jointly learns both tasks via Contextualized Knowledge Distillation (ConKD). We introduce two versions of ConKD: hard gate and soft gate. The former selectively gates between two task-specific teachers, while the latter integrates knowledge from both teachers. Our gates are computed on-the-fly in a context-specific manner, facilitating flexible integration of relevant knowledge. Extensive experiments demonstrate that our single model significantly improves recommendation performance while enhancing fluency, and achieves comparable results in terms of diversity. | [
"Jung, Yeongseo",
"Jung, Eunseo",
"Chen, Lei"
] | Towards a Unified Conversational Recommendation System: Multi-task Learning via Contextualized Knowledge Distillation | emnlp-main.840 | 2310.18119 | [
""
] | https://huggingface.co/papers/2310.18119 | 0 | 0 | 0 | 3 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.841.bib | https://aclanthology.org/2023.emnlp-main.841/ | @inproceedings{chan-etal-2023-clair,
title = "{CLAIR}: Evaluating Image Captions with Large Language Models",
author = "Chan, David and
Petryk, Suzanne and
Gonzalez, Joseph and
Darrell, Trevor and
Canny, John",
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.841",
doi = "10.18653/v1/2023.emnlp-main.841",
pages = "13638--13646",
abstract = "The evaluation of machine-generated image captions poses an interesting yet persistent challenge. Effective evaluation measures must consider numerous dimensions of similarity, including semantic relevance, visual structure, object interactions, caption diversity, and specificity. Existing highly-engineered measures attempt to capture specific aspects, but fall short in providing a holistic score that aligns closely with human judgments. Here, we propose CLAIR, a novel method that leverages the zero-shot language modeling capabilities of large language models (LLMs) to evaluate candidate captions. In our evaluations, CLAIR demonstrates a stronger correlation with human judgments of caption quality compared to existing measures. Notably, on Flickr8K-Expert, CLAIR achieves relative correlation improvements over SPICE of 39.6{\%} and over image-augmented methods such as RefCLIP-S of 18.3{\%}. Moreover, CLAIR provides noisily interpretable results by allowing the language model to identify the underlying reasoning behind its assigned score.",
}
| The evaluation of machine-generated image captions poses an interesting yet persistent challenge. Effective evaluation measures must consider numerous dimensions of similarity, including semantic relevance, visual structure, object interactions, caption diversity, and specificity. Existing highly-engineered measures attempt to capture specific aspects, but fall short in providing a holistic score that aligns closely with human judgments. Here, we propose CLAIR, a novel method that leverages the zero-shot language modeling capabilities of large language models (LLMs) to evaluate candidate captions. In our evaluations, CLAIR demonstrates a stronger correlation with human judgments of caption quality compared to existing measures. Notably, on Flickr8K-Expert, CLAIR achieves relative correlation improvements over SPICE of 39.6{\%} and over image-augmented methods such as RefCLIP-S of 18.3{\%}. Moreover, CLAIR provides noisily interpretable results by allowing the language model to identify the underlying reasoning behind its assigned score. | [
"Chan, David",
"Petryk, Suzanne",
"Gonzalez, Joseph",
"Darrell, Trevor",
"Canny, John"
] | CLAIR: Evaluating Image Captions with Large Language Models | emnlp-main.841 | 2310.12971 | [
""
] | https://huggingface.co/papers/2310.12971 | 1 | 0 | 0 | 5 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.842.bib | https://aclanthology.org/2023.emnlp-main.842/ | @inproceedings{li-etal-2023-mope,
title = "{M}o{P}e: Model Perturbation based Privacy Attacks on Language Models",
author = "Li, Marvin and
Wang, Jason and
Wang, Jeffrey and
Neel, Seth",
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.842",
doi = "10.18653/v1/2023.emnlp-main.842",
pages = "13647--13660",
abstract = "Recent work has shown that Large Language Models (LLMs) can unintentionally leak sensitive information present in their training data. In this paper, we present Model Perturbations (MoPe), a new method to identify with high confidence if a given text is in the training data of a pre-trained language model, given white-box access to the models parameters. MoPe adds noise to the model in parameter space and measures the drop in log-likelihood at a given point $x$, a statistic we show approximates the trace of the Hessian matrix with respect to model parameters. Across language models ranging from 70M to 12B parameters, we show that MoPe is more effective than existing loss-based attacks and recently proposed perturbation-based methods. We also examine the role of training point order and model size in attack success, and empirically demonstrate that MoPe accurately approximate the trace of the Hessian in practice. Our results show that the loss of a point alone is insufficient to determine extractability{---}there are training points we can recover using our method that have average loss. This casts some doubt on prior works that use the loss of a point as evidence of memorization or unlearning.",
}
| Recent work has shown that Large Language Models (LLMs) can unintentionally leak sensitive information present in their training data. In this paper, we present Model Perturbations (MoPe), a new method to identify with high confidence if a given text is in the training data of a pre-trained language model, given white-box access to the models parameters. MoPe adds noise to the model in parameter space and measures the drop in log-likelihood at a given point $x$, a statistic we show approximates the trace of the Hessian matrix with respect to model parameters. Across language models ranging from 70M to 12B parameters, we show that MoPe is more effective than existing loss-based attacks and recently proposed perturbation-based methods. We also examine the role of training point order and model size in attack success, and empirically demonstrate that MoPe accurately approximate the trace of the Hessian in practice. Our results show that the loss of a point alone is insufficient to determine extractability{---}there are training points we can recover using our method that have average loss. This casts some doubt on prior works that use the loss of a point as evidence of memorization or unlearning. | [
"Li, Marvin",
"Wang, Jason",
"Wang, Jeffrey",
"Neel, Seth"
] | MoPe: Model Perturbation based Privacy Attacks on Language Models | emnlp-main.842 | 2310.14369 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.843.bib | https://aclanthology.org/2023.emnlp-main.843/ | @inproceedings{bitton-etal-2023-q2d,
title = "q2d: Turning Questions into Dialogs to Teach Models How to Search",
author = "Bitton, Yonatan and
Cohen-Ganor, Shlomi and
Hakimi, Ido and
Lewenberg, Yoad and
Aharoni, Roee and
Weinreb, Enav",
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.843",
doi = "10.18653/v1/2023.emnlp-main.843",
pages = "13661--13676",
abstract = "One of the exciting capabilities of recent language models for dialog is their ability to independently search for relevant information to ground a given dialog response. However, obtaining training data to teach models how to issue search queries is time and resource consuming. In this work, we propose $q2d$: an automatic data generation pipeline that generates information-seeking dialogs from questions. We prompt a large language model (PaLM) to create conversational versions of question answering datasets, and use it to improve query generation models that communicate with external search APIs to ground dialog responses. Unlike previous approaches which relied on human written dialogs with search queries, our method allows to automatically generate query-based grounded dialogs with better control and scale. Our experiments demonstrate that: (1) For query generation on the QReCC dataset, models trained on our synthetically-generated data achieve 90{\%}-97{\%} of the performance of models trained on the human-generated data; (2) We can successfully generate data for training dialog models in new domains without any existing dialog data as demonstrated on the multi-hop MuSiQue and Bamboogle QA datasets. (3) We perform a thorough analysis of the generated dialogs showing that humans find them of high quality and struggle to distinguish them from human-written dialogs.",
}
| One of the exciting capabilities of recent language models for dialog is their ability to independently search for relevant information to ground a given dialog response. However, obtaining training data to teach models how to issue search queries is time and resource consuming. In this work, we propose $q2d$: an automatic data generation pipeline that generates information-seeking dialogs from questions. We prompt a large language model (PaLM) to create conversational versions of question answering datasets, and use it to improve query generation models that communicate with external search APIs to ground dialog responses. Unlike previous approaches which relied on human written dialogs with search queries, our method allows to automatically generate query-based grounded dialogs with better control and scale. Our experiments demonstrate that: (1) For query generation on the QReCC dataset, models trained on our synthetically-generated data achieve 90{\%}-97{\%} of the performance of models trained on the human-generated data; (2) We can successfully generate data for training dialog models in new domains without any existing dialog data as demonstrated on the multi-hop MuSiQue and Bamboogle QA datasets. (3) We perform a thorough analysis of the generated dialogs showing that humans find them of high quality and struggle to distinguish them from human-written dialogs. | [
"Bitton, Yonatan",
"Cohen-Ganor, Shlomi",
"Hakimi, Ido",
"Lewenberg, Yoad",
"Aharoni, Roee",
"Weinreb, Enav"
] | q2d: Turning Questions into Dialogs to Teach Models How to Search | emnlp-main.843 | 2304.14318 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.844.bib | https://aclanthology.org/2023.emnlp-main.844/ | @inproceedings{kim-etal-2023-aligning,
title = "Aligning Large Language Models through Synthetic Feedback",
author = "Kim, Sungdong and
Bae, Sanghwan and
Shin, Jamin and
Kang, Soyoung and
Kwak, Donghyun and
Yoo, Kang and
Seo, Minjoon",
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.844",
doi = "10.18653/v1/2023.emnlp-main.844",
pages = "13677--13700",
abstract = "Aligning large language models (LLMs) to human values has become increasingly important as it enables sophisticated steering of LLMs. However, it requires significant human demonstrations and feedback or distillation from proprietary LLMs such as ChatGPT. In this work, we propose a novel alignment learning framework with synthetic feedback not dependent on extensive human annotations and proprietary LLMs. First, we perform reward modeling (RM) with synthetic feedback by contrasting responses from vanilla LLMs with various sizes and prompts. Then, we use the RM to simulate high-quality demonstrations to train a supervised policy and further optimize the model with reinforcement learning. Our resulting model, Aligned Language Model with Synthetic Training dataset (ALMoST), outperforms recent open-sourced models, which are trained on the outputs of InstructGPT or human-annotated demonstrations, in alignment benchmarks. In human evaluation, our model is preferred to Alpaca and Dolly-v2, 55.0{\%} and 58.5{\%} of the time, respectively. Further analyses demonstrate the efficacy and importance of synthetic feedback in our framework.",
}
| Aligning large language models (LLMs) to human values has become increasingly important as it enables sophisticated steering of LLMs. However, it requires significant human demonstrations and feedback or distillation from proprietary LLMs such as ChatGPT. In this work, we propose a novel alignment learning framework with synthetic feedback not dependent on extensive human annotations and proprietary LLMs. First, we perform reward modeling (RM) with synthetic feedback by contrasting responses from vanilla LLMs with various sizes and prompts. Then, we use the RM to simulate high-quality demonstrations to train a supervised policy and further optimize the model with reinforcement learning. Our resulting model, Aligned Language Model with Synthetic Training dataset (ALMoST), outperforms recent open-sourced models, which are trained on the outputs of InstructGPT or human-annotated demonstrations, in alignment benchmarks. In human evaluation, our model is preferred to Alpaca and Dolly-v2, 55.0{\%} and 58.5{\%} of the time, respectively. Further analyses demonstrate the efficacy and importance of synthetic feedback in our framework. | [
"Kim, Sungdong",
"Bae, Sanghwan",
"Shin, Jamin",
"Kang, Soyoung",
"Kwak, Donghyun",
"Yoo, Kang",
"Seo, Minjoon"
] | Aligning Large Language Models through Synthetic Feedback | emnlp-main.844 | 2305.13735 | [
"https://github.com/naver-ai/almost"
] | https://huggingface.co/papers/2305.13735 | 2 | 1 | 0 | 7 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.845.bib | https://aclanthology.org/2023.emnlp-main.845/ | @inproceedings{baranov-etal-2023-told,
title = "You Told Me That Joke Twice: A Systematic Investigation of Transferability and Robustness of Humor Detection Models",
author = "Baranov, Alexander and
Kniazhevsky, Vladimir and
Braslavski, Pavel",
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.845",
doi = "10.18653/v1/2023.emnlp-main.845",
pages = "13701--13715",
abstract = {In this study, we focus on automatic humor detection, a highly relevant task for conversational AI. To date, there are several English datasets for this task, but little research on how models trained on them generalize and behave in the wild. To fill this gap, we carefully analyze existing datasets, train RoBERTa-based and Na{\"\i}ve Bayes classifiers on each of them, and test on the rest. Training and testing on the same dataset yields good results, but the transferability of the models varies widely. Models trained on datasets with jokes from different sources show better transferability, while the amount of training data has a smaller impact. The behavior of the models on out-of-domain data is unstable, suggesting that some of the models overfit, while others learn non-specific humor characteristics. An adversarial attack shows that models trained on pun datasets are less robust. We also evaluate the sense of humor of the chatGPT and Flan-UL2 models in a zero-shot scenario. The LLMs demonstrate competitive results on humor datasets and a more stable behavior on out-of-domain data. We believe that the obtained results will facilitate the development of new datasets and evaluation methodologies in the field of computational humor. We{'}ve made all the data from the study and the trained models publicly available at https://github.com/Humor-Research/Humor-detection.},
}
| In this study, we focus on automatic humor detection, a highly relevant task for conversational AI. To date, there are several English datasets for this task, but little research on how models trained on them generalize and behave in the wild. To fill this gap, we carefully analyze existing datasets, train RoBERTa-based and Na{\"\i}ve Bayes classifiers on each of them, and test on the rest. Training and testing on the same dataset yields good results, but the transferability of the models varies widely. Models trained on datasets with jokes from different sources show better transferability, while the amount of training data has a smaller impact. The behavior of the models on out-of-domain data is unstable, suggesting that some of the models overfit, while others learn non-specific humor characteristics. An adversarial attack shows that models trained on pun datasets are less robust. We also evaluate the sense of humor of the chatGPT and Flan-UL2 models in a zero-shot scenario. The LLMs demonstrate competitive results on humor datasets and a more stable behavior on out-of-domain data. We believe that the obtained results will facilitate the development of new datasets and evaluation methodologies in the field of computational humor. We{'}ve made all the data from the study and the trained models publicly available at https://github.com/Humor-Research/Humor-detection. | [
"Baranov, Alex",
"er",
"Kniazhevsky, Vladimir",
"Braslavski, Pavel"
] | You Told Me That Joke Twice: A Systematic Investigation of Transferability and Robustness of Humor Detection Models | emnlp-main.845 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Oral |
|
https://aclanthology.org/2023.emnlp-main.846.bib | https://aclanthology.org/2023.emnlp-main.846/ | @inproceedings{zhang-etal-2023-reading,
title = "Reading Order Matters: Information Extraction from Visually-rich Documents by Token Path Prediction",
author = "Zhang, Chong and
Guo, Ya and
Tu, Yi and
Chen, Huan and
Tang, Jinyang and
Zhu, Huijia and
Zhang, Qi and
Gui, Tao",
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.846",
doi = "10.18653/v1/2023.emnlp-main.846",
pages = "13716--13730",
abstract = "Recent advances in multimodal pre-trained models have significantly improved information extraction from visually-rich documents (VrDs), in which named entity recognition (NER) is treated as a sequence-labeling task of predicting the BIO entity tags for tokens, following the typical setting of NLP. However, BIO-tagging scheme relies on the correct order of model inputs, which is not guaranteed in real-world NER on scanned VrDs where text are recognized and arranged by OCR systems. Such reading order issue hinders the accurate marking of entities by BIO-tagging scheme, making it impossible for sequence-labeling methods to predict correct named entities. To address the reading order issue, we introduce Token Path Prediction (TPP), a simple prediction head to predict entity mentions as token sequences within documents. Alternative to token classification, TPP models the document layout as a complete directed graph of tokens, and predicts token paths within the graph as entities. For better evaluation of VrD-NER systems, we also propose two revised benchmark datasets of NER on scanned documents which can reflect real-world scenarios. Experiment results demonstrate the effectiveness of our method, and suggest its potential to be a universal solution to various information extraction tasks on documents.",
}
| Recent advances in multimodal pre-trained models have significantly improved information extraction from visually-rich documents (VrDs), in which named entity recognition (NER) is treated as a sequence-labeling task of predicting the BIO entity tags for tokens, following the typical setting of NLP. However, BIO-tagging scheme relies on the correct order of model inputs, which is not guaranteed in real-world NER on scanned VrDs where text are recognized and arranged by OCR systems. Such reading order issue hinders the accurate marking of entities by BIO-tagging scheme, making it impossible for sequence-labeling methods to predict correct named entities. To address the reading order issue, we introduce Token Path Prediction (TPP), a simple prediction head to predict entity mentions as token sequences within documents. Alternative to token classification, TPP models the document layout as a complete directed graph of tokens, and predicts token paths within the graph as entities. For better evaluation of VrD-NER systems, we also propose two revised benchmark datasets of NER on scanned documents which can reflect real-world scenarios. Experiment results demonstrate the effectiveness of our method, and suggest its potential to be a universal solution to various information extraction tasks on documents. | [
"Zhang, Chong",
"Guo, Ya",
"Tu, Yi",
"Chen, Huan",
"Tang, Jinyang",
"Zhu, Huijia",
"Zhang, Qi",
"Gui, Tao"
] | Reading Order Matters: Information Extraction from Visually-rich Documents by Token Path Prediction | emnlp-main.846 | 2310.11016 | [
"https://github.com/chongzhangfdu/tpp"
] | https://huggingface.co/papers/2310.11016 | 0 | 0 | 0 | 8 | [] | [] | [] | 1 | Oral |
https://aclanthology.org/2023.emnlp-main.847.bib | https://aclanthology.org/2023.emnlp-main.847/ | @inproceedings{wu-etal-2023-empower,
title = "Empower Nested {B}oolean Logic via Self-Supervised Curriculum Learning",
author = "Wu, Hongqiu and
Liu, Linfeng and
Zhao, Hai 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.847",
doi = "10.18653/v1/2023.emnlp-main.847",
pages = "13731--13742",
abstract = "Beyond the great cognitive powers showcased by language models, it is crucial to scrutinize whether their reasoning capabilities stem from strong generalization or merely exposure to relevant data. As opposed to constructing increasingly complex logic, this paper probes into the boolean logic, the root capability of a logical reasoner. We find that any pre-trained language models even including large language models only behave like a random selector in the face of multi-nested boolean logic, a task that humans can handle with ease. To empower language models with this fundamental capability, this paper proposes a new self-supervised learning method Curriculum Logical Reasoning (Clr), where we augment the training data with nested boolean logic chain step-by-step, and program the training from simpler logical patterns gradually to harder ones. This new training paradigm allows language models to effectively generalize to much harder and longer-hop logic, which can hardly be learned through naive training. Furthermore, we show that boolean logic is a great foundation for improving the subsequent general logical tasks.",
}
| Beyond the great cognitive powers showcased by language models, it is crucial to scrutinize whether their reasoning capabilities stem from strong generalization or merely exposure to relevant data. As opposed to constructing increasingly complex logic, this paper probes into the boolean logic, the root capability of a logical reasoner. We find that any pre-trained language models even including large language models only behave like a random selector in the face of multi-nested boolean logic, a task that humans can handle with ease. To empower language models with this fundamental capability, this paper proposes a new self-supervised learning method Curriculum Logical Reasoning (Clr), where we augment the training data with nested boolean logic chain step-by-step, and program the training from simpler logical patterns gradually to harder ones. This new training paradigm allows language models to effectively generalize to much harder and longer-hop logic, which can hardly be learned through naive training. Furthermore, we show that boolean logic is a great foundation for improving the subsequent general logical tasks. | [
"Wu, Hongqiu",
"Liu, Linfeng",
"Zhao, Hai",
"Zhang, Min"
] | Empower Nested Boolean Logic via Self-Supervised Curriculum Learning | emnlp-main.847 | 2310.05450 | [
"https://github.com/gingasan/boolkill"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.848.bib | https://aclanthology.org/2023.emnlp-main.848/ | @inproceedings{venkit-etal-2023-sentiment,
title = "The Sentiment Problem: A Critical Survey towards Deconstructing Sentiment Analysis",
author = "Venkit, Pranav and
Srinath, Mukund and
Gautam, Sanjana and
Venkatraman, Saranya and
Gupta, Vipul and
Passonneau, Rebecca and
Wilson, Shomir",
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.848",
doi = "10.18653/v1/2023.emnlp-main.848",
pages = "13743--13763",
abstract = "We conduct an inquiry into the sociotechnical aspects of sentiment analysis (SA) by critically examining 189 peer-reviewed papers on their applications, models, and datasets. Our investigation stems from the recognition that SA has become an integral component of diverse sociotechnical systems, exerting influence on both social and technical users. By delving into sociological and technological literature on sentiment, we unveil distinct conceptualizations of this term in domains such as finance, government, and medicine. Our study exposes a lack of explicit definitions and frameworks for characterizing sentiment, resulting in potential challenges and biases. To tackle this issue, we propose an ethics sheet encompassing critical inquiries to guide practitioners in ensuring equitable utilization of SA. Our findings underscore the significance of adopting an interdisciplinary approach to defining sentiment in SA and offer a pragmatic solution for its implementation.",
}
| We conduct an inquiry into the sociotechnical aspects of sentiment analysis (SA) by critically examining 189 peer-reviewed papers on their applications, models, and datasets. Our investigation stems from the recognition that SA has become an integral component of diverse sociotechnical systems, exerting influence on both social and technical users. By delving into sociological and technological literature on sentiment, we unveil distinct conceptualizations of this term in domains such as finance, government, and medicine. Our study exposes a lack of explicit definitions and frameworks for characterizing sentiment, resulting in potential challenges and biases. To tackle this issue, we propose an ethics sheet encompassing critical inquiries to guide practitioners in ensuring equitable utilization of SA. Our findings underscore the significance of adopting an interdisciplinary approach to defining sentiment in SA and offer a pragmatic solution for its implementation. | [
"Venkit, Pranav",
"Srinath, Mukund",
"Gautam, Sanjana",
"Venkatraman, Saranya",
"Gupta, Vipul",
"Passonneau, Rebecca",
"Wilson, Shomir"
] | The Sentiment Problem: A Critical Survey towards Deconstructing Sentiment Analysis | emnlp-main.848 | 2310.12318 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.849.bib | https://aclanthology.org/2023.emnlp-main.849/ | @inproceedings{zhong-etal-2023-poisoning,
title = "Poisoning Retrieval Corpora by Injecting Adversarial Passages",
author = "Zhong, Zexuan and
Huang, Ziqing and
Wettig, Alexander and
Chen, Danqi",
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.849",
doi = "10.18653/v1/2023.emnlp-main.849",
pages = "13764--13775",
abstract = "Dense retrievers have achieved state-of-the-art performance in various information retrieval tasks, but to what extent can they be safely deployed in real-world applications? In this work, we propose a novel attack for dense retrieval systems in which a malicious user generates a small number of adversarial passages by perturbing discrete tokens to maximize similarity with a provided set of training queries. When these adversarial passages are inserted into a large retrieval corpus, we show that this attack is highly effective in fooling these systems to retrieve them for queries that were not seen by the attacker. More surprisingly, these adversarial passages can directly generalize to out-of-domain queries and corpora with a high success attack rate {---} for instance, we find that 50 generated passages optimized on Natural Questions can mislead {\textgreater}94{\%} of questions posed in financial documents or online forums. We also benchmark and compare a range of state-of-the-art dense retrievers, both unsupervised and supervised. Although different systems exhibit varying levels of vulnerability, we show they can all be successfully attacked by injecting up to 500 passages, a small fraction compared to a retrieval corpus of millions of passages.",
}
| Dense retrievers have achieved state-of-the-art performance in various information retrieval tasks, but to what extent can they be safely deployed in real-world applications? In this work, we propose a novel attack for dense retrieval systems in which a malicious user generates a small number of adversarial passages by perturbing discrete tokens to maximize similarity with a provided set of training queries. When these adversarial passages are inserted into a large retrieval corpus, we show that this attack is highly effective in fooling these systems to retrieve them for queries that were not seen by the attacker. More surprisingly, these adversarial passages can directly generalize to out-of-domain queries and corpora with a high success attack rate {---} for instance, we find that 50 generated passages optimized on Natural Questions can mislead {\textgreater}94{\%} of questions posed in financial documents or online forums. We also benchmark and compare a range of state-of-the-art dense retrievers, both unsupervised and supervised. Although different systems exhibit varying levels of vulnerability, we show they can all be successfully attacked by injecting up to 500 passages, a small fraction compared to a retrieval corpus of millions of passages. | [
"Zhong, Zexuan",
"Huang, Ziqing",
"Wettig, Alex",
"er",
"Chen, Danqi"
] | Poisoning Retrieval Corpora by Injecting Adversarial Passages | emnlp-main.849 | 2310.19156 | [
"https://github.com/princeton-nlp/corpus-poisoning"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.850.bib | https://aclanthology.org/2023.emnlp-main.850/ | @inproceedings{liu-etal-2023-dada,
title = "{DADA}: Dialect Adaptation via Dynamic Aggregation of Linguistic Rules",
author = "Liu, Yanchen and
Held, William 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.850",
doi = "10.18653/v1/2023.emnlp-main.850",
pages = "13776--13793",
abstract = "Existing large language models (LLMs) that mainly focus on Standard American English (SAE) often lead to significantly worse performance when being applied to other English dialects. While existing mitigations tackle discrepancies for individual target dialects, they assume access to high-accuracy dialect identification systems. The boundaries between dialects are inherently flexible, making it difficult to categorize language into discrete predefined categories. In this paper, we propose DADA (Dialect Adaptation via Dynamic Aggregation), a modular approach to imbue SAE-trained models with multi-dialectal robustness by composing adapters which handle specific linguistic features. The compositional architecture of DADA allows for both targeted adaptation to specific dialect variants and simultaneous adaptation to various dialects. We show that DADA is effective for both single task and instruction finetuned language models, offering an extensible and interpretable framework for adapting existing LLMs to different English dialects.",
}
| Existing large language models (LLMs) that mainly focus on Standard American English (SAE) often lead to significantly worse performance when being applied to other English dialects. While existing mitigations tackle discrepancies for individual target dialects, they assume access to high-accuracy dialect identification systems. The boundaries between dialects are inherently flexible, making it difficult to categorize language into discrete predefined categories. In this paper, we propose DADA (Dialect Adaptation via Dynamic Aggregation), a modular approach to imbue SAE-trained models with multi-dialectal robustness by composing adapters which handle specific linguistic features. The compositional architecture of DADA allows for both targeted adaptation to specific dialect variants and simultaneous adaptation to various dialects. We show that DADA is effective for both single task and instruction finetuned language models, offering an extensible and interpretable framework for adapting existing LLMs to different English dialects. | [
"Liu, Yanchen",
"Held, William",
"Yang, Diyi"
] | DADA: Dialect Adaptation via Dynamic Aggregation of Linguistic Rules | emnlp-main.850 | 2305.13406 | [
"https://github.com/salt-nlp/dada"
] | https://huggingface.co/papers/2305.13406 | 1 | 0 | 0 | 3 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.851.bib | https://aclanthology.org/2023.emnlp-main.851/ | @inproceedings{ma-etal-2023-clustering,
title = "Clustering Pseudo Language Family in Multilingual Translation Models with Fisher Information Matrix",
author = "Ma, Xinyu and
Liu, Xuebo 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.851",
doi = "10.18653/v1/2023.emnlp-main.851",
pages = "13794--13804",
abstract = "In multilingual translation research, the comprehension and utilization of language families are of paramount importance. Nevertheless, clustering languages based solely on their ancestral families can yield suboptimal results due to variations in the datasets employed during the model{'}s training phase. To mitigate this challenge, we introduce an innovative method that leverages the fisher information matrix (FIM) to cluster language families, anchored on the multilingual translation model{'}s characteristics. We hypothesize that language pairs with similar effects on model parameters exhibit a considerable degree of linguistic congruence and should thus be grouped cohesively. This concept has led us to define pseudo language families. We provide an in-depth discussion regarding the inception and application of these pseudo language families. Empirical evaluations reveal that employing these pseudo language families enhances performance over conventional language families in adapting a multilingual translation model to unfamiliar language pairs. The proposed methodology may also be extended to scenarios requiring language similarity measurements. The source code and associated scripts can be accessed at https://github.com/ecoli-hit/PseudoFamily.",
}
| In multilingual translation research, the comprehension and utilization of language families are of paramount importance. Nevertheless, clustering languages based solely on their ancestral families can yield suboptimal results due to variations in the datasets employed during the model{'}s training phase. To mitigate this challenge, we introduce an innovative method that leverages the fisher information matrix (FIM) to cluster language families, anchored on the multilingual translation model{'}s characteristics. We hypothesize that language pairs with similar effects on model parameters exhibit a considerable degree of linguistic congruence and should thus be grouped cohesively. This concept has led us to define pseudo language families. We provide an in-depth discussion regarding the inception and application of these pseudo language families. Empirical evaluations reveal that employing these pseudo language families enhances performance over conventional language families in adapting a multilingual translation model to unfamiliar language pairs. The proposed methodology may also be extended to scenarios requiring language similarity measurements. The source code and associated scripts can be accessed at https://github.com/ecoli-hit/PseudoFamily. | [
"Ma, Xinyu",
"Liu, Xuebo",
"Zhang, Min"
] | Clustering Pseudo Language Family in Multilingual Translation Models with Fisher Information Matrix | emnlp-main.851 | 2312.02820 | [
"https://github.com/ecoli-hit/pseudofamily"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.852.bib | https://aclanthology.org/2023.emnlp-main.852/ | @inproceedings{xue-etal-2023-unifying,
title = "Unifying Discrete and Continuous Representations for Unsupervised Paraphrase Generation",
author = "Xue, Mingfeng and
Liu, Dayiheng and
Lei, Wenqiang and
Fu, Jie and
Lan, Jian and
Li, Mei and
Yang, Baosong and
Xie, Jun and
Zhang, Yidan and
Peng, Dezhong 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.852",
doi = "10.18653/v1/2023.emnlp-main.852",
pages = "13805--13822",
abstract = "Unsupervised paraphrase generation is a challenging task that benefits a variety of downstream NLP applications. Current unsupervised methods for paraphrase generation typically employ round-trip translation or denoising, which require translation corpus and result in paraphrases overly similar to the original sentences in surface structure. Most of these methods lack explicit control over the similarity between the original and generated sentences, and the entities are also less correctly kept. To obviate the reliance on translation data and prompt greater variations in surface structure, we propose a self-supervised pseudo-data construction method that generates diverse pseudo-paraphrases in distinct surface structures for a given sentence. To control the similarity and generate accurate entities, we propose an unsupervised paraphrasing model that encodes the sentence meaning and the entities with discrete and continuous variables, respectively. The similarity can be controlled by sampling discrete variables and the entities are kept substantially accurate due to the specific modeling of entities using continuous variables. Experimental results on two benchmark datasets demonstrate the advantages of our pseudo-data construction method compared to round-trip translation, and the superiority of our paraphrasing model over the state-of-the-art unsupervised methods.",
}
| Unsupervised paraphrase generation is a challenging task that benefits a variety of downstream NLP applications. Current unsupervised methods for paraphrase generation typically employ round-trip translation or denoising, which require translation corpus and result in paraphrases overly similar to the original sentences in surface structure. Most of these methods lack explicit control over the similarity between the original and generated sentences, and the entities are also less correctly kept. To obviate the reliance on translation data and prompt greater variations in surface structure, we propose a self-supervised pseudo-data construction method that generates diverse pseudo-paraphrases in distinct surface structures for a given sentence. To control the similarity and generate accurate entities, we propose an unsupervised paraphrasing model that encodes the sentence meaning and the entities with discrete and continuous variables, respectively. The similarity can be controlled by sampling discrete variables and the entities are kept substantially accurate due to the specific modeling of entities using continuous variables. Experimental results on two benchmark datasets demonstrate the advantages of our pseudo-data construction method compared to round-trip translation, and the superiority of our paraphrasing model over the state-of-the-art unsupervised methods. | [
"Xue, Mingfeng",
"Liu, Dayiheng",
"Lei, Wenqiang",
"Fu, Jie",
"Lan, Jian",
"Li, Mei",
"Yang, Baosong",
"Xie, Jun",
"Zhang, Yidan",
"Peng, Dezhong",
"Lv, Jiancheng"
] | Unifying Discrete and Continuous Representations for Unsupervised Paraphrase Generation | emnlp-main.852 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.853.bib | https://aclanthology.org/2023.emnlp-main.853/ | @inproceedings{gao-etal-2023-benefits,
title = "The Benefits of Label-Description Training for Zero-Shot Text Classification",
author = "Gao, Lingyu and
Ghosh, Debanjan and
Gimpel, Kevin",
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.853",
doi = "10.18653/v1/2023.emnlp-main.853",
pages = "13823--13844",
abstract = "Pretrained language models have improved zero-shot text classification by allowing the transfer of semantic knowledge from the training data in order to classify among specific label sets in downstream tasks. We propose a simple way to further improve zero-shot accuracies with minimal effort. We curate small finetuning datasets intended to describe the labels for a task. Unlike typical finetuning data, which has texts annotated with labels, our data simply describes the labels in language, e.g., using a few related terms, dictionary/encyclopedia entries, and short templates. Across a range of topic and sentiment datasets, our method is more accurate than zero-shot by 17-19{\%} absolute. It is also more robust to choices required for zero-shot classification, such as patterns for prompting the model to classify and mappings from labels to tokens in the model{'}s vocabulary. Furthermore, since our data merely describes the labels but does not use input texts, finetuning on it yields a model that performs strongly on multiple text domains for a given label set, even improving over few-shot out-of-domain classification in multiple settings.",
}
| Pretrained language models have improved zero-shot text classification by allowing the transfer of semantic knowledge from the training data in order to classify among specific label sets in downstream tasks. We propose a simple way to further improve zero-shot accuracies with minimal effort. We curate small finetuning datasets intended to describe the labels for a task. Unlike typical finetuning data, which has texts annotated with labels, our data simply describes the labels in language, e.g., using a few related terms, dictionary/encyclopedia entries, and short templates. Across a range of topic and sentiment datasets, our method is more accurate than zero-shot by 17-19{\%} absolute. It is also more robust to choices required for zero-shot classification, such as patterns for prompting the model to classify and mappings from labels to tokens in the model{'}s vocabulary. Furthermore, since our data merely describes the labels but does not use input texts, finetuning on it yields a model that performs strongly on multiple text domains for a given label set, even improving over few-shot out-of-domain classification in multiple settings. | [
"Gao, Lingyu",
"Ghosh, Debanjan",
"Gimpel, Kevin"
] | The Benefits of Label-Description Training for Zero-Shot Text Classification | emnlp-main.853 | 2305.02239 | [
"https://github.com/lingyugao/labeldesctraining"
] | https://huggingface.co/papers/2305.02239 | 1 | 0 | 0 | 3 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.854.bib | https://aclanthology.org/2023.emnlp-main.854/ | @inproceedings{salesky-etal-2023-multilingual,
title = "Multilingual Pixel Representations for Translation and Effective Cross-lingual Transfer",
author = "Salesky, Elizabeth and
Verma, Neha and
Koehn, Philipp and
Post, Matt",
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.854",
doi = "10.18653/v1/2023.emnlp-main.854",
pages = "13845--13861",
abstract = "We introduce and demonstrate how to effectively train multilingual machine translation models with pixel representations. We experiment with two different data settings with a variety of language and script coverage, demonstrating improved performance compared to subword embeddings. We explore various properties of pixel representations such as parameter sharing within and across scripts to better understand where they lead to positive transfer. We observe that these properties not only enable seamless cross-lingual transfer to unseen scripts, but make pixel representations more data-efficient than alternatives such as vocabulary expansion. We hope this work contributes to more extensible multilingual models for all languages and scripts.",
}
| We introduce and demonstrate how to effectively train multilingual machine translation models with pixel representations. We experiment with two different data settings with a variety of language and script coverage, demonstrating improved performance compared to subword embeddings. We explore various properties of pixel representations such as parameter sharing within and across scripts to better understand where they lead to positive transfer. We observe that these properties not only enable seamless cross-lingual transfer to unseen scripts, but make pixel representations more data-efficient than alternatives such as vocabulary expansion. We hope this work contributes to more extensible multilingual models for all languages and scripts. | [
"Salesky, Elizabeth",
"Verma, Neha",
"Koehn, Philipp",
"Post, Matt"
] | Multilingual Pixel Representations for Translation and Effective Cross-lingual Transfer | emnlp-main.854 | 2305.14280 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Oral |
|
https://aclanthology.org/2023.emnlp-main.855.bib | https://aclanthology.org/2023.emnlp-main.855/ | @inproceedings{albanyan-etal-2023-finding,
title = "Finding Authentic Counterhate Arguments: A Case Study with Public Figures",
author = "Albanyan, Abdullah and
Hassan, Ahmed and
Blanco, Eduardo",
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.855",
doi = "10.18653/v1/2023.emnlp-main.855",
pages = "13862--13876",
abstract = "We explore authentic counterhate arguments for online hateful content toward individuals. Previous efforts are limited to counterhate to fight against hateful content toward groups. Thus, we present a corpus of 54,816 hateful tweet-paragraph pairs, where the paragraphs are candidate counterhate arguments. The counterhate arguments are retrieved from 2,500 online articles from multiple sources. We propose a methodology that assures the authenticity of the counter argument and its specificity to the individual of interest. We show that finding arguments in online articles is an efficient alternative to counterhate generation approaches that may hallucinate unsupported arguments. We also present linguistic insights on the language used in counterhate arguments. Experimental results show promising results. It is more challenging, however, to identify counterhate arguments for hateful content toward individuals not included in the training set.",
}
| We explore authentic counterhate arguments for online hateful content toward individuals. Previous efforts are limited to counterhate to fight against hateful content toward groups. Thus, we present a corpus of 54,816 hateful tweet-paragraph pairs, where the paragraphs are candidate counterhate arguments. The counterhate arguments are retrieved from 2,500 online articles from multiple sources. We propose a methodology that assures the authenticity of the counter argument and its specificity to the individual of interest. We show that finding arguments in online articles is an efficient alternative to counterhate generation approaches that may hallucinate unsupported arguments. We also present linguistic insights on the language used in counterhate arguments. Experimental results show promising results. It is more challenging, however, to identify counterhate arguments for hateful content toward individuals not included in the training set. | [
"Albanyan, Abdullah",
"Hassan, Ahmed",
"Blanco, Eduardo"
] | Finding Authentic Counterhate Arguments: A Case Study with Public Figures | emnlp-main.855 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.856.bib | https://aclanthology.org/2023.emnlp-main.856/ | @inproceedings{cheng-etal-2023-edit,
title = "Can We Edit Multimodal Large Language Models?",
author = "Cheng, Siyuan and
Tian, Bozhong and
Liu, Qingbin and
Chen, Xi and
Wang, Yongheng 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.856",
doi = "10.18653/v1/2023.emnlp-main.856",
pages = "13877--13888",
abstract = "In this paper, we focus on editing multimodal Large Language Models (LLMs). Compared to editing single-modal LLMs, multimodal model editing is more challenging, which demands a higher level of scrutiny and careful consideration in the editing process. To facilitate research in this area, we construct a new benchmark, dubbed MMEdit, for editing multimodal LLMs and establishing a suite of innovative metrics for evaluation. We conduct comprehensive experiments involving various model editing baselines and analyze the impact of editing different components for multimodal LLMs. Empirically, we notice that previous baselines can implement editing multimodal LLMs to some extent, but the effect is still barely satisfactory, indicating the potential difficulty of this task. We hope that our work can provide the NLP community with insights.",
}
| In this paper, we focus on editing multimodal Large Language Models (LLMs). Compared to editing single-modal LLMs, multimodal model editing is more challenging, which demands a higher level of scrutiny and careful consideration in the editing process. To facilitate research in this area, we construct a new benchmark, dubbed MMEdit, for editing multimodal LLMs and establishing a suite of innovative metrics for evaluation. We conduct comprehensive experiments involving various model editing baselines and analyze the impact of editing different components for multimodal LLMs. Empirically, we notice that previous baselines can implement editing multimodal LLMs to some extent, but the effect is still barely satisfactory, indicating the potential difficulty of this task. We hope that our work can provide the NLP community with insights. | [
"Cheng, Siyuan",
"Tian, Bozhong",
"Liu, Qingbin",
"Chen, Xi",
"Wang, Yongheng",
"Chen, Huajun",
"Zhang, Ningyu"
] | Can We Edit Multimodal Large Language Models? | emnlp-main.856 | 2310.08475 | [
"https://github.com/zjunlp/easyedit"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.857.bib | https://aclanthology.org/2023.emnlp-main.857/ | @inproceedings{hu-wan-2023-exploring,
title = "Exploring Discourse Structure in Document-level Machine Translation",
author = "Hu, Xinyu and
Wan, Xiaojun",
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.857",
doi = "10.18653/v1/2023.emnlp-main.857",
pages = "13889--13902",
abstract = "Neural machine translation has achieved great success in the past few years with the help of transformer architectures and large-scale bilingual corpora. However, when the source text gradually grows into an entire document, the performance of current methods for document-level machine translation (DocMT) is less satisfactory. Although the context is beneficial to the translation in general, it is difficult for traditional methods to utilize such long-range information. Previous studies on DocMT have concentrated on extra contents such as multiple surrounding sentences and input instances divided by a fixed length. We suppose that they ignore the structure inside the source text, which leads to under-utilization of the context. In this paper, we present a more sound paragraph-to-paragraph translation mode and explore whether discourse structure can improve DocMT. We introduce several methods from different perspectives, among which our RST-Att model with a multi-granularity attention mechanism based on the RST parsing tree works best. The experiments show that our method indeed utilizes discourse information and performs better than previous work.",
}
| Neural machine translation has achieved great success in the past few years with the help of transformer architectures and large-scale bilingual corpora. However, when the source text gradually grows into an entire document, the performance of current methods for document-level machine translation (DocMT) is less satisfactory. Although the context is beneficial to the translation in general, it is difficult for traditional methods to utilize such long-range information. Previous studies on DocMT have concentrated on extra contents such as multiple surrounding sentences and input instances divided by a fixed length. We suppose that they ignore the structure inside the source text, which leads to under-utilization of the context. In this paper, we present a more sound paragraph-to-paragraph translation mode and explore whether discourse structure can improve DocMT. We introduce several methods from different perspectives, among which our RST-Att model with a multi-granularity attention mechanism based on the RST parsing tree works best. The experiments show that our method indeed utilizes discourse information and performs better than previous work. | [
"Hu, Xinyu",
"Wan, Xiaojun"
] | Exploring Discourse Structure in Document-level Machine Translation | emnlp-main.857 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.858.bib | https://aclanthology.org/2023.emnlp-main.858/ | @inproceedings{zhang-etal-2023-clusterllm,
title = "{C}luster{LLM}: Large Language Models as a Guide for Text Clustering",
author = "Zhang, Yuwei and
Wang, Zihan and
Shang, 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.858",
doi = "10.18653/v1/2023.emnlp-main.858",
pages = "13903--13920",
abstract = "We introduce ClusterLLM, a novel text clustering framework that leverages feedback from an instruction-tuned large language model, such as ChatGPT. Compared with traditional unsupervised methods that builds upon {``}small{''} embedders, ClusterLLM exhibits two intriguing advantages: (1) it enjoys the emergent capability of LLM even if its embeddings are inaccessible; and (2) it understands the user{'}s preference on clustering through textual instruction and/or a few annotated data. First, we prompt ChatGPT for insights on clustering perspective by constructing hard triplet questions $<$does A better correspond to B than C$>$, where A, B and C are similar data points that belong to different clusters according to small embedder. We empirically show that this strategy is both effective for fine-tuning small embedder and cost-efficient to query ChatGPT. Second, we prompt ChatGPT for helps on clustering granularity by carefully designed pairwise questions $<$do A and B belong to the same category$>$, and tune the granularity from cluster hierarchies that is the most consistent with the ChatGPT answers. Extensive experiments on 14 datasets show that ClusterLLM consistently improves clustering quality, at an average cost of {\textasciitilde}{\$}0.6 per dataset.",
}
| We introduce ClusterLLM, a novel text clustering framework that leverages feedback from an instruction-tuned large language model, such as ChatGPT. Compared with traditional unsupervised methods that builds upon {``}small{''} embedders, ClusterLLM exhibits two intriguing advantages: (1) it enjoys the emergent capability of LLM even if its embeddings are inaccessible; and (2) it understands the user{'}s preference on clustering through textual instruction and/or a few annotated data. First, we prompt ChatGPT for insights on clustering perspective by constructing hard triplet questions $<$does A better correspond to B than C$>$, where A, B and C are similar data points that belong to different clusters according to small embedder. We empirically show that this strategy is both effective for fine-tuning small embedder and cost-efficient to query ChatGPT. Second, we prompt ChatGPT for helps on clustering granularity by carefully designed pairwise questions $<$do A and B belong to the same category$>$, and tune the granularity from cluster hierarchies that is the most consistent with the ChatGPT answers. Extensive experiments on 14 datasets show that ClusterLLM consistently improves clustering quality, at an average cost of {\textasciitilde}{\$}0.6 per dataset. | [
"Zhang, Yuwei",
"Wang, Zihan",
"Shang, Jingbo"
] | ClusterLLM: Large Language Models as a Guide for Text Clustering | emnlp-main.858 | 2305.14871 | [
"https://github.com/zhang-yu-wei/clusterllm"
] | https://huggingface.co/papers/2305.14871 | 0 | 0 | 0 | 3 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.859.bib | https://aclanthology.org/2023.emnlp-main.859/ | @inproceedings{zhou-etal-2023-codebertscore,
title = "{C}ode{BERTS}core: Evaluating Code Generation with Pretrained Models of Code",
author = "Zhou, Shuyan and
Alon, Uri and
Agarwal, Sumit and
Neubig, Graham",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.emnlp-main.859",
doi = "10.18653/v1/2023.emnlp-main.859",
pages = "13921--13937",
abstract = "Since the rise of neural natural-language-to-code models (NL$\rightarrow$Code) that can generate long expressions and statements rather than a single next-token, one of the major problems has been reliably evaluating their generated output. In this paper, we propose CodeBERTScore: an evaluation metric for code generation, which builds on BERTScore (Zhang et al., 2020). Instead of encoding only the generated tokens as in BERTScore, CodeBERTScore also encodes the natural language input preceding the generated code, thus modeling the consistency between the generated code and its given natural language context as well. We perform an extensive evaluation of CodeBERTScore across four programming languages. We find that CodeBERTScore achieves a higher correlation with human preference and with functional correctness than all existing metrics. That is, generated code that receives a higher score by CodeBERTScore is more likely to be preferred by humans, as well as to function correctly when executed. We release five language-specific pretrained models to use with our publicly available code. Our language-specific models have been downloaded more than **1,000,000** times from the Huggingface Hub. Our code and data are available at https://github.com/neulab/code-bert-score",
}
| Since the rise of neural natural-language-to-code models (NL$\rightarrow$Code) that can generate long expressions and statements rather than a single next-token, one of the major problems has been reliably evaluating their generated output. In this paper, we propose CodeBERTScore: an evaluation metric for code generation, which builds on BERTScore (Zhang et al., 2020). Instead of encoding only the generated tokens as in BERTScore, CodeBERTScore also encodes the natural language input preceding the generated code, thus modeling the consistency between the generated code and its given natural language context as well. We perform an extensive evaluation of CodeBERTScore across four programming languages. We find that CodeBERTScore achieves a higher correlation with human preference and with functional correctness than all existing metrics. That is, generated code that receives a higher score by CodeBERTScore is more likely to be preferred by humans, as well as to function correctly when executed. We release five language-specific pretrained models to use with our publicly available code. Our language-specific models have been downloaded more than **1,000,000** times from the Huggingface Hub. Our code and data are available at https://github.com/neulab/code-bert-score | [
"Zhou, Shuyan",
"Alon, Uri",
"Agarwal, Sumit",
"Neubig, Graham"
] | CodeBERTScore: Evaluating Code Generation with Pretrained Models of Code | emnlp-main.859 | 2302.05527 | [
"https://github.com/neulab/code-bert-score"
] | https://huggingface.co/papers/2302.05527 | 0 | 1 | 0 | 4 | [
"neulab/codebert-python",
"neulab/codebert-java",
"neulab/codebert-javascript",
"neulab/codebert-cpp",
"neulab/codebert-c"
] | [] | [
"mahdy151/neulab-codebert-python",
"aydenjf/neulab-codebert-python",
"throdriguesdev/neulab-codebert-python"
] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.860.bib | https://aclanthology.org/2023.emnlp-main.860/ | @inproceedings{huang-etal-2023-learn,
title = "Learn and Consolidate: Continual Adaptation for Zero-Shot and Multilingual Neural Machine Translation",
author = "Huang, Kaiyu and
Li, Peng and
Liu, Junpeng and
Sun, Maosong and
Liu, Yang",
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.860",
doi = "10.18653/v1/2023.emnlp-main.860",
pages = "13938--13951",
abstract = "Although existing multilingual neural machine translation (MNMT) models have demonstrated remarkable performance to handle multiple translation directions in a single model and achieved zero-shot translation between language pairs unseen in training, they still suffer from relatively poor translation qualities for some language pairs. A practical scenario is that how to continually update MNMT models for both supervised and zero-shot translations when limited new data arrives. To this end, we propose a two-stage approach that encourages original models to acquire language-agnostic multilingual representations from new data, and preserves the model architecture without introducing parameters. Experimental results and further analysis demonstrate that our method can efficiently improve performance of existing MNMT models in translation directions where they are initially weak, and mitigates the degeneration in the original well-performing translation directions, offering flexibility in the real-world scenario.",
}
| Although existing multilingual neural machine translation (MNMT) models have demonstrated remarkable performance to handle multiple translation directions in a single model and achieved zero-shot translation between language pairs unseen in training, they still suffer from relatively poor translation qualities for some language pairs. A practical scenario is that how to continually update MNMT models for both supervised and zero-shot translations when limited new data arrives. To this end, we propose a two-stage approach that encourages original models to acquire language-agnostic multilingual representations from new data, and preserves the model architecture without introducing parameters. Experimental results and further analysis demonstrate that our method can efficiently improve performance of existing MNMT models in translation directions where they are initially weak, and mitigates the degeneration in the original well-performing translation directions, offering flexibility in the real-world scenario. | [
"Huang, Kaiyu",
"Li, Peng",
"Liu, Junpeng",
"Sun, Maosong",
"Liu, Yang"
] | Learn and Consolidate: Continual Adaptation for Zero-Shot and Multilingual Neural Machine Translation | emnlp-main.860 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.861.bib | https://aclanthology.org/2023.emnlp-main.861/ | @inproceedings{mishra-etal-2023-e,
title = "e-{THERAPIST}: {I} suggest you to cultivate a mindset of positivity and nurture uplifting thoughts",
author = "Mishra, Kshitij and
Priya, Priyanshu and
Burja, Manisha and
Ekbal, Asif",
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.861",
doi = "10.18653/v1/2023.emnlp-main.861",
pages = "13952--13967",
abstract = "The shortage of therapists for mental health patients emphasizes the importance of globally accessible dialogue systems alleviating their issues. To have effective interpersonal psychotherapy, these systems must exhibit politeness and empathy when needed. However, these factors may vary as per the user{'}s gender, age, persona, and sentiment. Hence, in order to establish trust and provide a personalized cordial experience, it is essential that generated responses should be tailored to individual profiles and attributes. Focusing on this objective, we propose e-THERAPIST, a novel polite interpersonal psychotherapy dialogue system to address issues like depression, anxiety, schizophrenia, etc. We begin by curating a unique conversational dataset for psychotherapy, called PsyCon. It is annotated at two levels: (i) dialogue-level - including user{'}s profile information (gender, age, persona) and therapist{'}s psychotherapeutic approach; and (ii) utterance-level - encompassing user{'}s sentiment and therapist{'}s politeness, and interpersonal behaviour. Then, we devise a novel reward model to adapt correct polite interpersonal behaviour and use it to train e-THERAPIST on PsyCon employing NLPO loss. Our extensive empirical analysis validates the effectiveness of each component of the proposed e-THERAPIST demonstrating its potential impact in psychotherapy settings.",
}
| The shortage of therapists for mental health patients emphasizes the importance of globally accessible dialogue systems alleviating their issues. To have effective interpersonal psychotherapy, these systems must exhibit politeness and empathy when needed. However, these factors may vary as per the user{'}s gender, age, persona, and sentiment. Hence, in order to establish trust and provide a personalized cordial experience, it is essential that generated responses should be tailored to individual profiles and attributes. Focusing on this objective, we propose e-THERAPIST, a novel polite interpersonal psychotherapy dialogue system to address issues like depression, anxiety, schizophrenia, etc. We begin by curating a unique conversational dataset for psychotherapy, called PsyCon. It is annotated at two levels: (i) dialogue-level - including user{'}s profile information (gender, age, persona) and therapist{'}s psychotherapeutic approach; and (ii) utterance-level - encompassing user{'}s sentiment and therapist{'}s politeness, and interpersonal behaviour. Then, we devise a novel reward model to adapt correct polite interpersonal behaviour and use it to train e-THERAPIST on PsyCon employing NLPO loss. Our extensive empirical analysis validates the effectiveness of each component of the proposed e-THERAPIST demonstrating its potential impact in psychotherapy settings. | [
"Mishra, Kshitij",
"Priya, Priyanshu",
"Burja, Manisha",
"Ekbal, Asif"
] | e-THERAPIST: I suggest you to cultivate a mindset of positivity and nurture uplifting thoughts | emnlp-main.861 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Oral |
|
https://aclanthology.org/2023.emnlp-main.862.bib | https://aclanthology.org/2023.emnlp-main.862/ | @inproceedings{muhammad-etal-2023-afrisenti,
title = "{A}fri{S}enti: A {T}witter Sentiment Analysis Benchmark for {A}frican Languages",
author = "Muhammad, Shamsuddeen and
Abdulmumin, Idris and
Ayele, Abinew and
Ousidhoum, Nedjma and
Adelani, David and
Yimam, Seid and
Ahmad, Ibrahim and
Beloucif, Meriem and
Mohammad, Saif and
Ruder, Sebastian and
Hourrane, Oumaima and
Jorge, Alipio and
Brazdil, Pavel and
Ali, Felermino and
David, Davis and
Osei, Salomey and
Shehu-Bello, Bello and
Lawan, Falalu and
Gwadabe, Tajuddeen and
Rutunda, Samuel and
Belay, Tadesse Destaw and
Messelle, Wendimu and
Balcha, Hailu and
Chala, Sisay and
Gebremichael, Hagos and
Opoku, Bernard and
Arthur, Stephen",
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.862",
doi = "10.18653/v1/2023.emnlp-main.862",
pages = "13968--13981",
abstract = "Africa is home to over 2,000 languages from over six language families and has the highest linguistic diversity among all continents. This includes 75 languages with at least one million speakers each. Yet, there is little NLP research conducted on African languages. Crucial in enabling such research is the availability of high-quality annotated datasets. In this paper, we introduce AfriSenti, a sentiment analysis benchmark that contains a total of {\textgreater}110,000 tweets in 14 African languages (Amharic, Algerian Arabic, Hausa, Igbo, Kinyarwanda, Moroccan Arabic, Mozambican Portuguese, Nigerian Pidgin, Oromo, Swahili, Tigrinya, Twi, Xitsonga, and Yoruba) from four language families. The tweets were annotated by native speakers and used in the AfriSenti-SemEval shared task (with over 200 participants, see website: https://afrisenti-semeval.github.io). We describe the data collection methodology, annotation process, and the challenges we dealt with when curating each dataset. We further report baseline experiments conducted on the AfriSenti datasets and discuss their usefulness.",
}
| Africa is home to over 2,000 languages from over six language families and has the highest linguistic diversity among all continents. This includes 75 languages with at least one million speakers each. Yet, there is little NLP research conducted on African languages. Crucial in enabling such research is the availability of high-quality annotated datasets. In this paper, we introduce AfriSenti, a sentiment analysis benchmark that contains a total of {\textgreater}110,000 tweets in 14 African languages (Amharic, Algerian Arabic, Hausa, Igbo, Kinyarwanda, Moroccan Arabic, Mozambican Portuguese, Nigerian Pidgin, Oromo, Swahili, Tigrinya, Twi, Xitsonga, and Yoruba) from four language families. The tweets were annotated by native speakers and used in the AfriSenti-SemEval shared task (with over 200 participants, see website: https://afrisenti-semeval.github.io). We describe the data collection methodology, annotation process, and the challenges we dealt with when curating each dataset. We further report baseline experiments conducted on the AfriSenti datasets and discuss their usefulness. | [
"Muhammad, Shamsuddeen",
"Abdulmumin, Idris",
"Ayele, Abinew",
"Ousidhoum, Nedjma",
"Adelani, David",
"Yimam, Seid",
"Ahmad, Ibrahim",
"Beloucif, Meriem",
"Mohammad, Saif",
"Ruder, Sebastian",
"Hourrane, Oumaima",
"Jorge, Alipio",
"Brazdil, Pavel",
"Ali, Felermino",
"David, Davis",
"Osei, Salomey",
"Shehu-Bello, Bello",
"Lawan, Falalu",
"Gwadabe, Tajuddeen",
"Rutunda, Samuel",
"Belay, Tadesse Destaw",
"Messelle, Wendimu",
"Balcha, Hailu",
"Chala, Sisay",
"Gebremichael, Hagos",
"Opoku, Bernard",
"Arthur, Stephen"
] | AfriSenti: A Twitter Sentiment Analysis Benchmark for African Languages | emnlp-main.862 | 2302.08956 | [
"https://github.com/afrisenti-semeval/afrisent-semeval-2023"
] | https://huggingface.co/papers/2302.08956 | 0 | 0 | 0 | 26 | [
"Davlan/afrisenti-twitter-sentiment-afroxlmr-large"
] | [
"HausaNLP/AfriSenti-Twitter",
"shmuhammad/AfriSenti-twitter-sentiment"
] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.863.bib | https://aclanthology.org/2023.emnlp-main.863/ | @inproceedings{yang-etal-2023-quantifying,
title = "Quantifying Character Similarity with Vision Transformers",
author = "Yang, Xinmei and
Arora, Abhishek and
Jheng, Shao-Yu and
Dell, Melissa",
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.863",
doi = "10.18653/v1/2023.emnlp-main.863",
pages = "13982--13996",
abstract = "Record linkage is a bedrock of quantitative social science, as analyses often require linking data from multiple, noisy sources. Off-the-shelf string matching methods are widely used, as they are straightforward and cheap to implement and scale. Not all character substitutions are equally probable, and for some settings there are widely used handcrafted lists denoting which string substitutions are more likely, that improve the accuracy of string matching. However, such lists do not exist for many settings, skewing research with linked datasets towards a few high-resource contexts that are not representative of the diversity of human societies. This study develops an extensible way to measure character substitution costs for OCR{'}ed documents, by employing large-scale self-supervised training of vision transformers (ViT) with augmented digital fonts. For each language written with the CJK script, we contrastively learn a metric space where different augmentations of the same character are represented nearby. In this space, homoglyphic characters - those with similar appearance such as {``}O{''} and {``}0{''} - have similar vector representations. Using the cosine distance between characters{'} representations as the substitution cost in an edit distance matching algorithm significantly improves record linkage compared to other widely used string matching methods, as OCR errors tend to be homoglyphic in nature. Homoglyphs can plausibly capture character visual similarity across any script, including low-resource settings. We illustrate this by creating homoglyph sets for 3,000 year old ancient Chinese characters, which are highly pictorial. Fascinatingly, a ViT is able to capture relationships in how different abstract concepts were conceptualized by ancient societies, that have been noted in the archaeological literature.",
}
| Record linkage is a bedrock of quantitative social science, as analyses often require linking data from multiple, noisy sources. Off-the-shelf string matching methods are widely used, as they are straightforward and cheap to implement and scale. Not all character substitutions are equally probable, and for some settings there are widely used handcrafted lists denoting which string substitutions are more likely, that improve the accuracy of string matching. However, such lists do not exist for many settings, skewing research with linked datasets towards a few high-resource contexts that are not representative of the diversity of human societies. This study develops an extensible way to measure character substitution costs for OCR{'}ed documents, by employing large-scale self-supervised training of vision transformers (ViT) with augmented digital fonts. For each language written with the CJK script, we contrastively learn a metric space where different augmentations of the same character are represented nearby. In this space, homoglyphic characters - those with similar appearance such as {``}O{''} and {``}0{''} - have similar vector representations. Using the cosine distance between characters{'} representations as the substitution cost in an edit distance matching algorithm significantly improves record linkage compared to other widely used string matching methods, as OCR errors tend to be homoglyphic in nature. Homoglyphs can plausibly capture character visual similarity across any script, including low-resource settings. We illustrate this by creating homoglyph sets for 3,000 year old ancient Chinese characters, which are highly pictorial. Fascinatingly, a ViT is able to capture relationships in how different abstract concepts were conceptualized by ancient societies, that have been noted in the archaeological literature. | [
"Yang, Xinmei",
"Arora, Abhishek",
"Jheng, Shao-Yu",
"Dell, Melissa"
] | Quantifying Character Similarity with Vision Transformers | emnlp-main.863 | 2305.14672 | [
"https://github.com/dell-research-harvard/quantifying-character-similarity"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.864.bib | https://aclanthology.org/2023.emnlp-main.864/ | @inproceedings{deng-etal-2023-syllogistic,
title = "Syllogistic Reasoning for Legal Judgment Analysis",
author = "Deng, Wentao and
Pei, Jiahuan and
Kong, Keyi and
Chen, Zhe and
Wei, Furu and
Li, Yujun and
Ren, Zhaochun and
Chen, Zhumin and
Ren, Pengjie",
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.864",
doi = "10.18653/v1/2023.emnlp-main.864",
pages = "13997--14009",
abstract = "Legal judgment assistants are developing fast due to impressive progress of large language models (LLMs). However, people can hardly trust the results generated by a model without reliable analysis of legal judgement. For legal practitioners, it is common practice to utilize syllogistic reasoning to select and evaluate the arguments of the parties as part of the legal decision-making process. But the development of syllogistic reasoning for legal judgment analysis is hindered by the lack of resources: (1) there is no large-scale syllogistic reasoning dataset for legal judgment analysis, and (2) there is no set of established benchmarks for legal judgment analysis. In this paper, we construct and manually correct a syllogistic reasoning dataset for legal judgment analysis. The dataset contains 11,239 criminal cases which cover 4 criminal elements, 80 charges and 124 articles. We also select a set of large language models as benchmarks, and conduct a in-depth analysis of the capacity of their legal judgment analysis.",
}
| Legal judgment assistants are developing fast due to impressive progress of large language models (LLMs). However, people can hardly trust the results generated by a model without reliable analysis of legal judgement. For legal practitioners, it is common practice to utilize syllogistic reasoning to select and evaluate the arguments of the parties as part of the legal decision-making process. But the development of syllogistic reasoning for legal judgment analysis is hindered by the lack of resources: (1) there is no large-scale syllogistic reasoning dataset for legal judgment analysis, and (2) there is no set of established benchmarks for legal judgment analysis. In this paper, we construct and manually correct a syllogistic reasoning dataset for legal judgment analysis. The dataset contains 11,239 criminal cases which cover 4 criminal elements, 80 charges and 124 articles. We also select a set of large language models as benchmarks, and conduct a in-depth analysis of the capacity of their legal judgment analysis. | [
"Deng, Wentao",
"Pei, Jiahuan",
"Kong, Keyi",
"Chen, Zhe",
"Wei, Furu",
"Li, Yujun",
"Ren, Zhaochun",
"Chen, Zhumin",
"Ren, Pengjie"
] | Syllogistic Reasoning for Legal Judgment Analysis | emnlp-main.864 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Oral |
|
https://aclanthology.org/2023.emnlp-main.865.bib | https://aclanthology.org/2023.emnlp-main.865/ | @inproceedings{kim-etal-2023-improving,
title = "Improving Transformer-based Program Repair Model through False Behavior Diagnosis",
author = "Kim, Youngkyoung and
Kim, Misoo and
Lee, Eunseok",
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.865",
doi = "10.18653/v1/2023.emnlp-main.865",
pages = "14010--14023",
abstract = "Research on automated program repairs using transformer-based models has recently gained considerable attention. The comprehension of the erroneous behavior of a model enables the identification of its inherent capacity and provides insights for improvement. However, the current landscape of research on program repair models lacks an investigation of their false behavior. Thus, we propose a methodology for diagnosing and treating the false behaviors of transformer-based program repair models. Specifically, we propose 1) a behavior vector that quantifies the behavior of the model when it generates an output, 2) a behavior discriminator (BeDisc) that identifies false behaviors, and 3) two methods for false behavior treatment. Through a large-scale experiment on 55,562 instances employing four datasets and three models, the BeDisc exhibited a balanced accuracy of 86.6{\%} for false behavior classification. The first treatment, namely, early abortion, successfully eliminated 60.4{\%} of false behavior while preserving 97.4{\%} repair accuracy. Furthermore, the second treatment, namely, masked bypassing, resulted in an average improvement of 40.5{\%} in the top-1 repair accuracy. These experimental results demonstrated the importance of investigating false behaviors in program repair models.",
}
| Research on automated program repairs using transformer-based models has recently gained considerable attention. The comprehension of the erroneous behavior of a model enables the identification of its inherent capacity and provides insights for improvement. However, the current landscape of research on program repair models lacks an investigation of their false behavior. Thus, we propose a methodology for diagnosing and treating the false behaviors of transformer-based program repair models. Specifically, we propose 1) a behavior vector that quantifies the behavior of the model when it generates an output, 2) a behavior discriminator (BeDisc) that identifies false behaviors, and 3) two methods for false behavior treatment. Through a large-scale experiment on 55,562 instances employing four datasets and three models, the BeDisc exhibited a balanced accuracy of 86.6{\%} for false behavior classification. The first treatment, namely, early abortion, successfully eliminated 60.4{\%} of false behavior while preserving 97.4{\%} repair accuracy. Furthermore, the second treatment, namely, masked bypassing, resulted in an average improvement of 40.5{\%} in the top-1 repair accuracy. These experimental results demonstrated the importance of investigating false behaviors in program repair models. | [
"Kim, Youngkyoung",
"Kim, Misoo",
"Lee, Eunseok"
] | Improving Transformer-based Program Repair Model through False Behavior Diagnosis | emnlp-main.865 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.866.bib | https://aclanthology.org/2023.emnlp-main.866/ | @inproceedings{qi-etal-2023-sut,
title = "{SUT}: Active Defects Probing for Transcompiler Models",
author = "Qi, Mengnan and
Huang, Yufan and
Wang, Maoquan and
Yao, Yongqiang and
Liu, Zihan 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.866",
doi = "10.18653/v1/2023.emnlp-main.866",
pages = "14024--14034",
abstract = "Automatic Program translation has enormous application value and hence has been attracting significant interest from AI researchers. However, we observe that current program translation models still make elementary syntax errors, particularly, when the target language does not have syntax elements in the source language. Metrics like BLUE, CodeBLUE and computation accuracy may not expose these issues. In this paper we introduce a new metrics for programming language translation and these metrics address these basic syntax errors. We develop a novel active defects probing suite called Syntactic Unit Tests (SUT) which includes a highly interpretable evaluation harness for accuracy and test scoring. Experiments have shown that even powerful models like ChatGPT still make mistakes on these basic unit tests. Specifically, compared to previous program translation task evaluation dataset, its pass rate on our unit tests has decreased by 26.15{\%}. Further our evaluation harness reveal syntactic element errors in which these models exhibit deficiencies.",
}
| Automatic Program translation has enormous application value and hence has been attracting significant interest from AI researchers. However, we observe that current program translation models still make elementary syntax errors, particularly, when the target language does not have syntax elements in the source language. Metrics like BLUE, CodeBLUE and computation accuracy may not expose these issues. In this paper we introduce a new metrics for programming language translation and these metrics address these basic syntax errors. We develop a novel active defects probing suite called Syntactic Unit Tests (SUT) which includes a highly interpretable evaluation harness for accuracy and test scoring. Experiments have shown that even powerful models like ChatGPT still make mistakes on these basic unit tests. Specifically, compared to previous program translation task evaluation dataset, its pass rate on our unit tests has decreased by 26.15{\%}. Further our evaluation harness reveal syntactic element errors in which these models exhibit deficiencies. | [
"Qi, Mengnan",
"Huang, Yufan",
"Wang, Maoquan",
"Yao, Yongqiang",
"Liu, Zihan",
"Gu, Bin",
"Clement, Colin",
"Sundaresan, Neel"
] | SUT: Active Defects Probing for Transcompiler Models | emnlp-main.866 | 2310.14209 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.867.bib | https://aclanthology.org/2023.emnlp-main.867/ | @inproceedings{choi-etal-2023-kcts,
title = "{KCTS}: Knowledge-Constrained Tree Search Decoding with Token-Level Hallucination Detection",
author = "Choi, Sehyun and
Fang, Tianqing and
Wang, Zhaowei and
Song, Yangqiu",
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.867",
doi = "10.18653/v1/2023.emnlp-main.867",
pages = "14035--14053",
abstract = "Large Language Models (LLMs) have demonstrated remarkable human-level natural language generation capabilities. However, their potential to generate misinformation, often called the *hallucination* problem, poses a significant risk to their deployment. A common approach to address this issue is to retrieve relevant knowledge and fine-tune the LLM with the knowledge in its input. Unfortunately, this method incurs high training costs and may cause catastrophic forgetting for multi-tasking models. To overcome these limitations, we propose a knowledge-constrained decoding method called KCTS (Knowledge-Constrained Tree Search), which guides a frozen LM to generate text aligned with the reference knowledge at each decoding step using a knowledge classifier score and MCTS (Monte-Carlo Tree Search). To adapt the sequence-level knowledge classifier to token-level guidance, we also propose a novel token-level hallucination detection method called RIPA (Reward Inflection Point Approximation). Our empirical results on knowledge-grounded dialogue and abstractive summarization demonstrate the strength of KCTS as a plug-and-play, model-agnostic decoding method that can effectively reduce hallucinations in natural language generation.",
}
| Large Language Models (LLMs) have demonstrated remarkable human-level natural language generation capabilities. However, their potential to generate misinformation, often called the *hallucination* problem, poses a significant risk to their deployment. A common approach to address this issue is to retrieve relevant knowledge and fine-tune the LLM with the knowledge in its input. Unfortunately, this method incurs high training costs and may cause catastrophic forgetting for multi-tasking models. To overcome these limitations, we propose a knowledge-constrained decoding method called KCTS (Knowledge-Constrained Tree Search), which guides a frozen LM to generate text aligned with the reference knowledge at each decoding step using a knowledge classifier score and MCTS (Monte-Carlo Tree Search). To adapt the sequence-level knowledge classifier to token-level guidance, we also propose a novel token-level hallucination detection method called RIPA (Reward Inflection Point Approximation). Our empirical results on knowledge-grounded dialogue and abstractive summarization demonstrate the strength of KCTS as a plug-and-play, model-agnostic decoding method that can effectively reduce hallucinations in natural language generation. | [
"Choi, Sehyun",
"Fang, Tianqing",
"Wang, Zhaowei",
"Song, Yangqiu"
] | KCTS: Knowledge-Constrained Tree Search Decoding with Token-Level Hallucination Detection | emnlp-main.867 | 2310.09044 | [
"https://github.com/hkust-knowcomp/knowledge-constrained-decoding"
] | https://huggingface.co/papers/2310.09044 | 1 | 0 | 0 | 4 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.868.bib | https://aclanthology.org/2023.emnlp-main.868/ | @inproceedings{kothyari-etal-2023-crush4sql,
title = "{CRUSH}4{SQL}: Collective Retrieval Using Schema Hallucination For {T}ext2{SQL}",
author = "Kothyari, Mayank and
Dhingra, Dhruva and
Sarawagi, Sunita and
Chakrabarti, Soumen",
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.868",
doi = "10.18653/v1/2023.emnlp-main.868",
pages = "14054--14066",
abstract = "Existing Text-to-SQL generators require the entire schema to be encoded with the user text. This is expensive or impractical for large databases with tens of thousands of columns. Standard dense retrieval techniques are inadequate for schema subsetting of a large structured database, where the correct semantics of retrieval demands that we rank sets of schema elements rather than individual documents. In response, we propose a two-stage process for effective coverage during retrieval. First, we use an LLM to hallucinate a minimal DB schema that it deems adequate to answer the query. We use the hallucinated schema to retrieve a subset of the actual schema, by composing the results from multiple dense retrievals. Remarkably, hallucination {---} generally considered a nuisance {---} turns out to be actually useful as a bridging mechanism. Since no existing benchmarks exist for schema subsetting on large databases, we introduce two benchmarks: (1) A semi-synthetic dataset of 4502 schema elements, by taking a union of schema on the well-known SPIDER dataset, and (2) A real-life benchmark called SocialDB sourced from an actual large data warehouse comprising of 17844 schema elements. We show that our method leads to significantly higher recall than SOTA retrieval-based augmentation methods.",
}
| Existing Text-to-SQL generators require the entire schema to be encoded with the user text. This is expensive or impractical for large databases with tens of thousands of columns. Standard dense retrieval techniques are inadequate for schema subsetting of a large structured database, where the correct semantics of retrieval demands that we rank sets of schema elements rather than individual documents. In response, we propose a two-stage process for effective coverage during retrieval. First, we use an LLM to hallucinate a minimal DB schema that it deems adequate to answer the query. We use the hallucinated schema to retrieve a subset of the actual schema, by composing the results from multiple dense retrievals. Remarkably, hallucination {---} generally considered a nuisance {---} turns out to be actually useful as a bridging mechanism. Since no existing benchmarks exist for schema subsetting on large databases, we introduce two benchmarks: (1) A semi-synthetic dataset of 4502 schema elements, by taking a union of schema on the well-known SPIDER dataset, and (2) A real-life benchmark called SocialDB sourced from an actual large data warehouse comprising of 17844 schema elements. We show that our method leads to significantly higher recall than SOTA retrieval-based augmentation methods. | [
"Kothyari, Mayank",
"Dhingra, Dhruva",
"Sarawagi, Sunita",
"Chakrabarti, Soumen"
] | CRUSH4SQL: Collective Retrieval Using Schema Hallucination For Text2SQL | emnlp-main.868 | 2311.01173 | [
"https://github.com/imayk/crush4sql"
] | https://huggingface.co/papers/2311.01173 | 0 | 0 | 0 | 4 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.869.bib | https://aclanthology.org/2023.emnlp-main.869/ | @inproceedings{fanconi-etal-2023-reads,
title = "This Reads Like That: Deep Learning for Interpretable Natural Language Processing",
author = "Fanconi, Claudio and
Vandenhirtz, Moritz and
Husmann, Severin and
Vogt, Julia",
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.869",
doi = "10.18653/v1/2023.emnlp-main.869",
pages = "14067--14076",
abstract = "Prototype learning, a popular machine learning method designed for inherently interpretable decisions, leverages similarities to learned prototypes for classifying new data. While it is mainly applied in computer vision, in this work, we build upon prior research and further explore the extension of prototypical networks to natural language processing. We introduce a learned weighted similarity measure that enhances the similarity computation by focusing on informative dimensions of pre-trained sentence embeddings. Additionally, we propose a post-hoc explainability mechanism that extracts prediction-relevant words from both the prototype and input sentences. Finally, we empirically demonstrate that our proposed method not only improves predictive performance on the AG News and RT Polarity datasets over a previous prototype-based approach, but also improves the faithfulness of explanations compared to rationale-based recurrent convolutions.",
}
| Prototype learning, a popular machine learning method designed for inherently interpretable decisions, leverages similarities to learned prototypes for classifying new data. While it is mainly applied in computer vision, in this work, we build upon prior research and further explore the extension of prototypical networks to natural language processing. We introduce a learned weighted similarity measure that enhances the similarity computation by focusing on informative dimensions of pre-trained sentence embeddings. Additionally, we propose a post-hoc explainability mechanism that extracts prediction-relevant words from both the prototype and input sentences. Finally, we empirically demonstrate that our proposed method not only improves predictive performance on the AG News and RT Polarity datasets over a previous prototype-based approach, but also improves the faithfulness of explanations compared to rationale-based recurrent convolutions. | [
"Fanconi, Claudio",
"V",
"enhirtz, Moritz",
"Husmann, Severin",
"Vogt, Julia"
] | This Reads Like That: Deep Learning for Interpretable Natural Language Processing | emnlp-main.869 | 2310.17010 | [
"https://github.com/fanconic/this_reads_like_that"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.870.bib | https://aclanthology.org/2023.emnlp-main.870/ | @inproceedings{herzig-etal-2023-incorporating,
title = "Incorporating Structured Representations into Pretrained Vision {\&} Language Models Using Scene Graphs",
author = "Herzig, Roei and
Mendelson, Alon and
Karlinsky, Leonid and
Arbelle, Assaf and
Feris, Rogerio and
Darrell, Trevor and
Globerson, Amir",
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.870",
doi = "10.18653/v1/2023.emnlp-main.870",
pages = "14077--14098",
abstract = "Vision and language models (VLMs) have demonstrated remarkable zero-shot (ZS) performance in a variety of tasks. However, recent works have shown that even the best VLMs struggle to capture aspects of compositional scene understanding, such as object attributes, relations, and action states. In contrast, obtaining structured annotations, such as scene graphs (SGs), that could improve these models is time-consuming and costly, and thus cannot be used on a large scale. Here we ask whether small SG datasets can provide sufficient information for enhancing structured understanding of pretrained VLMs. We show that it is indeed possible to improve VLMs when learning from SGs by integrating components that incorporate structured information into both visual and textual representations. For the visual side, we incorporate a special {``}SG Component{''} in the image transformer trained to predict SG information, while for the textual side, we utilize SGs to generate fine-grained captions that highlight different compositional aspects of the scene. Our method improves the performance of several popular VLMs on multiple VL datasets with only a mild degradation in ZS capabilities.",
}
| Vision and language models (VLMs) have demonstrated remarkable zero-shot (ZS) performance in a variety of tasks. However, recent works have shown that even the best VLMs struggle to capture aspects of compositional scene understanding, such as object attributes, relations, and action states. In contrast, obtaining structured annotations, such as scene graphs (SGs), that could improve these models is time-consuming and costly, and thus cannot be used on a large scale. Here we ask whether small SG datasets can provide sufficient information for enhancing structured understanding of pretrained VLMs. We show that it is indeed possible to improve VLMs when learning from SGs by integrating components that incorporate structured information into both visual and textual representations. For the visual side, we incorporate a special {``}SG Component{''} in the image transformer trained to predict SG information, while for the textual side, we utilize SGs to generate fine-grained captions that highlight different compositional aspects of the scene. Our method improves the performance of several popular VLMs on multiple VL datasets with only a mild degradation in ZS capabilities. | [
"Herzig, Roei",
"Mendelson, Alon",
"Karlinsky, Leonid",
"Arbelle, Assaf",
"Feris, Rogerio",
"Darrell, Trevor",
"Globerson, Amir"
] | Incorporating Structured Representations into Pretrained Vision & Language Models Using Scene Graphs | emnlp-main.870 | 2305.06343 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.871.bib | https://aclanthology.org/2023.emnlp-main.871/ | @inproceedings{wu-etal-2023-tlm,
title = "{TLM}: Token-Level Masking for Transformers",
author = "Wu, Yangjun and
Fang, Kebin and
Zhang, Dongxiang and
Wang, Han and
Zhang, Hao and
Chen, Gang",
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.871",
doi = "10.18653/v1/2023.emnlp-main.871",
pages = "14099--14111",
abstract = "Structured dropout approaches, such as attention dropout and DropHead, have been investigated to regularize the multi-head attention mechanism in Transformers. In this paper, we propose a new regularization scheme based on token-level rather than structure-level to reduce overfitting. Specifically, we devise a novel Token-Level Masking (TLM) training strategy for Transformers to regularize the connections of self-attention, which consists of two masking techniques that are effective and easy to implement. The underlying idea is to manipulate the connections between tokens in the multi-head attention via masking, where the networks are forced to exploit partial neighbors{'} information to produce a meaningful representation. The generality and effectiveness of TLM are thoroughly evaluated via extensive experiments on 4 diversified NLP tasks across 18 datasets, including natural language understanding benchmark GLUE, ChineseGLUE, Chinese Grammatical Error Correction, and data-to-text generation. The results indicate that TLM can consistently outperform attention dropout and DropHead, e.g., it increases by 0.5 points relative to DropHead with BERT-large on GLUE. Moreover, TLM can establish a new record on the data-to-text benchmark Rotowire (18.93 BLEU). Our code will be publicly available at https://github.com/Young1993/tlm.",
}
| Structured dropout approaches, such as attention dropout and DropHead, have been investigated to regularize the multi-head attention mechanism in Transformers. In this paper, we propose a new regularization scheme based on token-level rather than structure-level to reduce overfitting. Specifically, we devise a novel Token-Level Masking (TLM) training strategy for Transformers to regularize the connections of self-attention, which consists of two masking techniques that are effective and easy to implement. The underlying idea is to manipulate the connections between tokens in the multi-head attention via masking, where the networks are forced to exploit partial neighbors{'} information to produce a meaningful representation. The generality and effectiveness of TLM are thoroughly evaluated via extensive experiments on 4 diversified NLP tasks across 18 datasets, including natural language understanding benchmark GLUE, ChineseGLUE, Chinese Grammatical Error Correction, and data-to-text generation. The results indicate that TLM can consistently outperform attention dropout and DropHead, e.g., it increases by 0.5 points relative to DropHead with BERT-large on GLUE. Moreover, TLM can establish a new record on the data-to-text benchmark Rotowire (18.93 BLEU). Our code will be publicly available at https://github.com/Young1993/tlm. | [
"Wu, Yangjun",
"Fang, Kebin",
"Zhang, Dongxiang",
"Wang, Han",
"Zhang, Hao",
"Chen, Gang"
] | TLM: Token-Level Masking for Transformers | emnlp-main.871 | 2310.18738 | [
"https://github.com/young1993/tlm"
] | https://huggingface.co/papers/2310.18738 | 0 | 1 | 0 | 6 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.872.bib | https://aclanthology.org/2023.emnlp-main.872/ | @inproceedings{liu-etal-2023-addressing,
title = "Addressing {NER} Annotation Noises with Uncertainty-Guided Tree-Structured {CRF}s",
author = "Liu, Jian and
Liu, Weichang and
Chen, Yufeng and
Xu, Jinan and
Zhao, Zhe",
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.872",
doi = "10.18653/v1/2023.emnlp-main.872",
pages = "14112--14123",
abstract = "Real-world named entity recognition (NER) datasets are notorious for their noisy nature, attributed to annotation errors, inconsistencies, and subjective interpretations. Such noises present a substantial challenge for traditional supervised learning methods. In this paper, we present a new and unified approach to tackle annotation noises for NER. Our method considers NER as a constituency tree parsing problem, utilizing a tree-structured Conditional Random Fields (CRFs) with uncertainty evaluation for integration. Through extensive experiments conducted on four real-world datasets, we demonstrate the effectiveness of our model in addressing both partial and incorrect annotation errors. Remarkably, our model exhibits superb performance even in extreme scenarios with 90{\%} annotation noise.",
}
| Real-world named entity recognition (NER) datasets are notorious for their noisy nature, attributed to annotation errors, inconsistencies, and subjective interpretations. Such noises present a substantial challenge for traditional supervised learning methods. In this paper, we present a new and unified approach to tackle annotation noises for NER. Our method considers NER as a constituency tree parsing problem, utilizing a tree-structured Conditional Random Fields (CRFs) with uncertainty evaluation for integration. Through extensive experiments conducted on four real-world datasets, we demonstrate the effectiveness of our model in addressing both partial and incorrect annotation errors. Remarkably, our model exhibits superb performance even in extreme scenarios with 90{\%} annotation noise. | [
"Liu, Jian",
"Liu, Weichang",
"Chen, Yufeng",
"Xu, Jinan",
"Zhao, Zhe"
] | Addressing NER Annotation Noises with Uncertainty-Guided Tree-Structured CRFs | emnlp-main.872 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.873.bib | https://aclanthology.org/2023.emnlp-main.873/ | @inproceedings{piergentili-etal-2023-hi,
title = "Hi Guys or Hi Folks? Benchmarking Gender-Neutral Machine Translation with the {G}e{NTE} Corpus",
author = "Piergentili, Andrea and
Savoldi, Beatrice and
Fucci, Dennis and
Negri, Matteo and
Bentivogli, Luisa",
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.873",
doi = "10.18653/v1/2023.emnlp-main.873",
pages = "14124--14140",
abstract = "Gender inequality is embedded in our communication practices and perpetuated in translation technologies. This becomes particularly apparent when translating into grammatical gender languages, where machine translation (MT) often defaults to masculine and stereotypical representations by making undue binary gender assumptions. Our work addresses the rising demand for inclusive language by focusing head-on on gender-neutral translation from English to Italian. We start from the essentials: proposing a dedicated benchmark and exploring automated evaluation methods. First, we introduce GeNTE, a natural, bilingual test set for gender-neutral translation, whose creation was informed by a survey on the perception and use of neutral language. Based on GeNTE, we then overview existing reference-based evaluation approaches, highlight their limits, and propose a reference-free method more suitable to assess gender-neutral translation.",
}
| Gender inequality is embedded in our communication practices and perpetuated in translation technologies. This becomes particularly apparent when translating into grammatical gender languages, where machine translation (MT) often defaults to masculine and stereotypical representations by making undue binary gender assumptions. Our work addresses the rising demand for inclusive language by focusing head-on on gender-neutral translation from English to Italian. We start from the essentials: proposing a dedicated benchmark and exploring automated evaluation methods. First, we introduce GeNTE, a natural, bilingual test set for gender-neutral translation, whose creation was informed by a survey on the perception and use of neutral language. Based on GeNTE, we then overview existing reference-based evaluation approaches, highlight their limits, and propose a reference-free method more suitable to assess gender-neutral translation. | [
"Piergentili, Andrea",
"Savoldi, Beatrice",
"Fucci, Dennis",
"Negri, Matteo",
"Bentivogli, Luisa"
] | Hi Guys or Hi Folks? Benchmarking Gender-Neutral Machine Translation with the GeNTE Corpus | emnlp-main.873 | 2310.05294 | [
"https://github.com/hlt-mt/fbk-neutr-eval"
] | https://huggingface.co/papers/2310.05294 | 2 | 0 | 0 | 5 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.874.bib | https://aclanthology.org/2023.emnlp-main.874/ | @inproceedings{costa-jussa-etal-2023-multilingual,
title = "Multilingual Holistic Bias: Extending Descriptors and Patterns to Unveil Demographic Biases in Languages at Scale",
author = "Costa-juss{\`a}, Marta and
Andrews, Pierre and
Smith, Eric and
Hansanti, Prangthip and
Ropers, Christophe and
Kalbassi, Elahe and
Gao, Cynthia and
Licht, Daniel and
Wood, Carleigh",
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.874",
doi = "10.18653/v1/2023.emnlp-main.874",
pages = "14141--14156",
abstract = "We introduce a multilingual extension of the HolisticBias dataset, the largest English template-based taxonomy of textual people references: Multilingual HolisticBias. This extension consists of 20,459 sentences in 50 languages distributed across 13 demographic axes. Source sentences are built from combinations of 118 demographic descriptors and three patterns, excluding nonsensical combinations. Multilingual translations include alternatives for gendered languages that cover gendered translations when there is ambiguity in English. Our dataset is intended to uncover demographic imbalances and be the tool to quantify mitigations towards them. Our initial findings show that translation quality for EN-to-XX translations is an average of almost 8 spBLEU better when evaluating with the masculine human reference compared to feminine. In the opposite direction, XX-to-EN, we compare the robustness of the model when the source input only differs in gender (masculine or feminine) and masculine translations are an average of almost 4 spBLEU better than feminine. When embedding sentences to a joint multilingual sentence representations space, we find that for most languages masculine translations are significantly closer to the English neutral sentences when embedded.",
}
| We introduce a multilingual extension of the HolisticBias dataset, the largest English template-based taxonomy of textual people references: Multilingual HolisticBias. This extension consists of 20,459 sentences in 50 languages distributed across 13 demographic axes. Source sentences are built from combinations of 118 demographic descriptors and three patterns, excluding nonsensical combinations. Multilingual translations include alternatives for gendered languages that cover gendered translations when there is ambiguity in English. Our dataset is intended to uncover demographic imbalances and be the tool to quantify mitigations towards them. Our initial findings show that translation quality for EN-to-XX translations is an average of almost 8 spBLEU better when evaluating with the masculine human reference compared to feminine. In the opposite direction, XX-to-EN, we compare the robustness of the model when the source input only differs in gender (masculine or feminine) and masculine translations are an average of almost 4 spBLEU better than feminine. When embedding sentences to a joint multilingual sentence representations space, we find that for most languages masculine translations are significantly closer to the English neutral sentences when embedded. | [
"Costa-juss{\\`a}, Marta",
"Andrews, Pierre",
"Smith, Eric",
"Hansanti, Prangthip",
"Ropers, Christophe",
"Kalbassi, Elahe",
"Gao, Cynthia",
"Licht, Daniel",
"Wood, Carleigh"
] | Multilingual Holistic Bias: Extending Descriptors and Patterns to Unveil Demographic Biases in Languages at Scale | emnlp-main.874 | 2305.13198 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.875.bib | https://aclanthology.org/2023.emnlp-main.875/ | @inproceedings{song-etal-2023-globalbench,
title = "{G}lobal{B}ench: A Benchmark for Global Progress in Natural Language Processing",
author = "Song, Yueqi and
Khanuja, Simran and
Liu, Pengfei and
Faisal, Fahim and
Ostapenko, Alissa and
Winata, Genta and
Aji, Alham Fikri and
Cahyawijaya, Samuel and
Tsvetkov, Yulia and
Anastasopoulos, Antonios and
Neubig, Graham",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.emnlp-main.875",
doi = "10.18653/v1/2023.emnlp-main.875",
pages = "14157--14171",
abstract = "Despite the major advances in NLP, significant disparities in NLP system performance across languages still exist. Arguably, these are due to uneven resource allocation and sub-optimal incentives to work on less resourced languages. To track and further incentivize the global development of equitable language technology, we introduce GlobalBench. Prior multilingual benchmarks are static and have focused on a limited number of tasks and languages. In contrast, GlobalBench is an ever-expanding collection that aims to dynamically track progress on all NLP datasets in all languages. Rather than solely measuring accuracy, GlobalBench also tracks the estimated per-speaker utility and equity of technology across all languages, providing a multi-faceted view of how language technology is serving people of the world. Furthermore, GlobalBench is designed to identify the most under-served languages, and rewards research efforts directed towards those languages. At present, the most under-served languages are the ones with a relatively high population, but nonetheless overlooked by composite multilingual benchmarks (like Punjabi, Portuguese, and Wu Chinese). Currently, GlobalBench covers 966 datasets in 190 languages, and has 1,128 system submissions spanning 62 languages.",
}
| Despite the major advances in NLP, significant disparities in NLP system performance across languages still exist. Arguably, these are due to uneven resource allocation and sub-optimal incentives to work on less resourced languages. To track and further incentivize the global development of equitable language technology, we introduce GlobalBench. Prior multilingual benchmarks are static and have focused on a limited number of tasks and languages. In contrast, GlobalBench is an ever-expanding collection that aims to dynamically track progress on all NLP datasets in all languages. Rather than solely measuring accuracy, GlobalBench also tracks the estimated per-speaker utility and equity of technology across all languages, providing a multi-faceted view of how language technology is serving people of the world. Furthermore, GlobalBench is designed to identify the most under-served languages, and rewards research efforts directed towards those languages. At present, the most under-served languages are the ones with a relatively high population, but nonetheless overlooked by composite multilingual benchmarks (like Punjabi, Portuguese, and Wu Chinese). Currently, GlobalBench covers 966 datasets in 190 languages, and has 1,128 system submissions spanning 62 languages. | [
"Song, Yueqi",
"Khanuja, Simran",
"Liu, Pengfei",
"Faisal, Fahim",
"Ostapenko, Alissa",
"Winata, Genta",
"Aji, Alham Fikri",
"Cahyawijaya, Samuel",
"Tsvetkov, Yulia",
"Anastasopoulos, Antonios",
"Neubig, Graham"
] | GlobalBench: A Benchmark for Global Progress in Natural Language Processing | emnlp-main.875 | 2305.14716 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.876.bib | https://aclanthology.org/2023.emnlp-main.876/ | @inproceedings{pi-etal-2023-detgpt,
title = "{D}et{GPT}: Detect What You Need via Reasoning",
author = "Pi, Renjie and
Gao, Jiahui and
Diao, Shizhe and
Pan, Rui and
Dong, Hanze and
Zhang, Jipeng and
Yao, Lewei and
Han, Jianhua and
Xu, Hang and
Kong, Lingpeng and
Zhang, Tong",
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.876",
doi = "10.18653/v1/2023.emnlp-main.876",
pages = "14172--14189",
abstract = "In recent years, the field of computer vision has seen significant advancements thanks to the development of large language models (LLMs). These models have enabled more effective and sophisticated interactions between humans and machines, paving the way for novel techniques that blur the lines between human and machine intelligence. In this paper, we introduce a new paradigm for object detection that we call reasoning-based object detection. Unlike conventional object detection methods that rely on specific object names, our approach enables users to interact with the system using natural language instructions, allowing for a higher level of interactivity. Our proposed method, called DetGPT, leverages state-of-the-art multi-modal models and open-vocabulary object detectors to perform reasoning within the context of the user{'}s instructions and the visual scene. This enables DetGPT to automatically locate the object of interest based on the user{'}s expressed desires, even if the object is not explicitly mentioned. For instance, if a user expresses a desire for a cold beverage, DetGPT can analyze the image, identify a fridge, and use its knowledge of typical fridge contents to locate the beverage. This flexibility makes our system applicable across a wide range of fields, from robotics and automation to autonomous driving. Overall, our proposed paradigm and DetGPT demonstrate the potential for more sophisticated and intuitive interactions between humans and machines. We hope that our proposed paradigm and approach will provide inspiration to the community and open the door to more interactive and versatile object detection systems.",
}
| In recent years, the field of computer vision has seen significant advancements thanks to the development of large language models (LLMs). These models have enabled more effective and sophisticated interactions between humans and machines, paving the way for novel techniques that blur the lines between human and machine intelligence. In this paper, we introduce a new paradigm for object detection that we call reasoning-based object detection. Unlike conventional object detection methods that rely on specific object names, our approach enables users to interact with the system using natural language instructions, allowing for a higher level of interactivity. Our proposed method, called DetGPT, leverages state-of-the-art multi-modal models and open-vocabulary object detectors to perform reasoning within the context of the user{'}s instructions and the visual scene. This enables DetGPT to automatically locate the object of interest based on the user{'}s expressed desires, even if the object is not explicitly mentioned. For instance, if a user expresses a desire for a cold beverage, DetGPT can analyze the image, identify a fridge, and use its knowledge of typical fridge contents to locate the beverage. This flexibility makes our system applicable across a wide range of fields, from robotics and automation to autonomous driving. Overall, our proposed paradigm and DetGPT demonstrate the potential for more sophisticated and intuitive interactions between humans and machines. We hope that our proposed paradigm and approach will provide inspiration to the community and open the door to more interactive and versatile object detection systems. | [
"Pi, Renjie",
"Gao, Jiahui",
"Diao, Shizhe",
"Pan, Rui",
"Dong, Hanze",
"Zhang, Jipeng",
"Yao, Lewei",
"Han, Jianhua",
"Xu, Hang",
"Kong, Lingpeng",
"Zhang, Tong"
] | DetGPT: Detect What You Need via Reasoning | emnlp-main.876 | 2305.14167 | [
""
] | https://huggingface.co/papers/2305.14167 | 3 | 0 | 0 | 11 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.877.bib | https://aclanthology.org/2023.emnlp-main.877/ | @inproceedings{kassner-etal-2023-language,
title = "Language Models with Rationality",
author = "Kassner, Nora and
Tafjord, Oyvind and
Sabharwal, Ashish and
Richardson, Kyle and
Schuetze, Hinrich and
Clark, Peter",
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.877",
doi = "10.18653/v1/2023.emnlp-main.877",
pages = "14190--14201",
abstract = "While large language models (LLMs) are proficient at question-answering (QA), it is not always clear how (or even if) an answer follows from their latent {``}beliefs{''}. This lack of interpretability is a growing impediment to widespread use of LLMs. To address this, our goals are to make model beliefs and their inferential relationships explicit, and to resolve inconsistencies that may exist, so that answers are supported by interpretable chains of reasoning drawn from a consistent network of beliefs. Our approach, which we call REFLEX, is to add a **rational, self-reflecting layer** on top of the LLM. First, given a question, we construct a **belief graph** using a backward-chaining process to materialize relevant model beliefs (including beliefs about answer candidates) and their inferential relationships. Second, we identify and minimize contradictions in that graph using a formal constraint reasoner. We find that REFLEX significantly improves consistency (by 8{\%}-11{\%} absolute) without harming overall answer accuracy, resulting in answers supported by faithful chains of reasoning drawn from a more consistent belief system. This suggests a new style of system architecture in which an LLM extended with a rational layer can provide an interpretable window into system beliefs, add a systematic reasoning capability, and repair latent inconsistencies present in the LLM.",
}
| While large language models (LLMs) are proficient at question-answering (QA), it is not always clear how (or even if) an answer follows from their latent {``}beliefs{''}. This lack of interpretability is a growing impediment to widespread use of LLMs. To address this, our goals are to make model beliefs and their inferential relationships explicit, and to resolve inconsistencies that may exist, so that answers are supported by interpretable chains of reasoning drawn from a consistent network of beliefs. Our approach, which we call REFLEX, is to add a **rational, self-reflecting layer** on top of the LLM. First, given a question, we construct a **belief graph** using a backward-chaining process to materialize relevant model beliefs (including beliefs about answer candidates) and their inferential relationships. Second, we identify and minimize contradictions in that graph using a formal constraint reasoner. We find that REFLEX significantly improves consistency (by 8{\%}-11{\%} absolute) without harming overall answer accuracy, resulting in answers supported by faithful chains of reasoning drawn from a more consistent belief system. This suggests a new style of system architecture in which an LLM extended with a rational layer can provide an interpretable window into system beliefs, add a systematic reasoning capability, and repair latent inconsistencies present in the LLM. | [
"Kassner, Nora",
"Tafjord, Oyvind",
"Sabharwal, Ashish",
"Richardson, Kyle",
"Schuetze, Hinrich",
"Clark, Peter"
] | Language Models with Rationality | emnlp-main.877 | 2407.06112 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.878.bib | https://aclanthology.org/2023.emnlp-main.878/ | @inproceedings{liao-etal-2023-self,
title = "Self-Improvement of Non-autoregressive Model via Sequence-Level Distillation",
author = "Liao, Yusheng and
Jiang, Shuyang and
Li, Yiqi and
Wang, Yu and
Wang, Yanfeng",
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.878",
doi = "10.18653/v1/2023.emnlp-main.878",
pages = "14202--14212",
abstract = "Although Non-autoregressive Transformer (NAT) models have achieved great success in terms of fast inference speed, this speedup comes with a performance drop due to the inherent \textit{multi-modality} problem of the NAT model. Previous works commonly alleviate this problem by replacing the target side of the raw data with distilled data generated by Autoregressive Transformer (AT) models. However, the multi-modality problem in the distilled data is still significant and thus limits further improvement of the NAT models. In this paper, we propose a method called Sequence-Level Self-Distillation (SLSD), which aims to generate distilled data by the NAT model itself, eliminating the need for additional teacher networks. Furthermore, SLSD can adapt to different NAT models without precise adjustments since the self-distilled data is generated from the same types of NAT models. We conduct extensive experiments on WMT14 EN$\leftrightarrow$DE and WMT16 EN$\leftrightarrow$RO and choose four classic NAT models as the backbones to validate the generality and effectiveness of SLSD. The results show that our approach can consistently improve all models on both raw data and distilled data without sacrificing the inference speed.",
}
| Although Non-autoregressive Transformer (NAT) models have achieved great success in terms of fast inference speed, this speedup comes with a performance drop due to the inherent \textit{multi-modality} problem of the NAT model. Previous works commonly alleviate this problem by replacing the target side of the raw data with distilled data generated by Autoregressive Transformer (AT) models. However, the multi-modality problem in the distilled data is still significant and thus limits further improvement of the NAT models. In this paper, we propose a method called Sequence-Level Self-Distillation (SLSD), which aims to generate distilled data by the NAT model itself, eliminating the need for additional teacher networks. Furthermore, SLSD can adapt to different NAT models without precise adjustments since the self-distilled data is generated from the same types of NAT models. We conduct extensive experiments on WMT14 EN$\leftrightarrow$DE and WMT16 EN$\leftrightarrow$RO and choose four classic NAT models as the backbones to validate the generality and effectiveness of SLSD. The results show that our approach can consistently improve all models on both raw data and distilled data without sacrificing the inference speed. | [
"Liao, Yusheng",
"Jiang, Shuyang",
"Li, Yiqi",
"Wang, Yu",
"Wang, Yanfeng"
] | Self-Improvement of Non-autoregressive Model via Sequence-Level Distillation | emnlp-main.878 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.879.bib | https://aclanthology.org/2023.emnlp-main.879/ | @inproceedings{zhang-choi-2023-mitigating,
title = "Mitigating Temporal Misalignment by Discarding Outdated Facts",
author = "Zhang, Michael and
Choi, Eunsol",
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.879",
doi = "10.18653/v1/2023.emnlp-main.879",
pages = "14213--14226",
abstract = "While large language models are able to retain vast amounts of world knowledge seen during pretraining, such knowledge is prone to going out of date and is nontrivial to update. Furthermore, these models are often used under temporal misalignment, tasked with answering questions about the present, despite having only been trained on data collected in the past. To mitigate the effects of temporal misalignment, we propose fact duration prediction: the task of predicting how long a given fact will remain true. In our experiments, we demonstrate that identifying which facts are prone to rapid change can help models avoid reciting outdated information and determine which predictions require seeking out up-to-date knowledge sources. We also show how modeling fact duration improves calibration for knowledge-intensive tasks, such as open-retrieval question answering, under temporal misalignment, by discarding volatile facts.",
}
| While large language models are able to retain vast amounts of world knowledge seen during pretraining, such knowledge is prone to going out of date and is nontrivial to update. Furthermore, these models are often used under temporal misalignment, tasked with answering questions about the present, despite having only been trained on data collected in the past. To mitigate the effects of temporal misalignment, we propose fact duration prediction: the task of predicting how long a given fact will remain true. In our experiments, we demonstrate that identifying which facts are prone to rapid change can help models avoid reciting outdated information and determine which predictions require seeking out up-to-date knowledge sources. We also show how modeling fact duration improves calibration for knowledge-intensive tasks, such as open-retrieval question answering, under temporal misalignment, by discarding volatile facts. | [
"Zhang, Michael",
"Choi, Eunsol"
] | Mitigating Temporal Misalignment by Discarding Outdated Facts | emnlp-main.879 | 2305.14824 | [
"https://github.com/mikejqzhang/mitigating_misalignment"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.880.bib | https://aclanthology.org/2023.emnlp-main.880/ | @inproceedings{hogan-etal-2023-open,
title = "Open-world Semi-supervised Generalized Relation Discovery Aligned in a Real-world Setting",
author = "Hogan, William and
Li, Jiacheng and
Shang, 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.880",
doi = "10.18653/v1/2023.emnlp-main.880",
pages = "14227--14242",
abstract = "Open-world Relation Extraction (OpenRE) has recently garnered significant attention. However, existing approaches tend to oversimplify the problem by assuming that all instances of unlabeled data belong to novel classes, thereby limiting the practicality of these methods. We argue that the OpenRE setting should be more aligned with the characteristics of real-world data. Specifically, we propose two key improvements: (a) unlabeled data should encompass known and novel classes, including negative instances; and (b) the set of novel classes should represent long-tail relation types. Furthermore, we observe that popular relations can often be implicitly inferred through specific patterns, while long-tail relations tend to be explicitly expressed. Motivated by these insights, we present a method called KNoRD (Known and Novel Relation Discovery), which effectively classifies explicitly and implicitly expressed relations from known and novel classes within unlabeled data. Experimental evaluations on several Open-world RE benchmarks demonstrate that KNoRD consistently outperforms other existing methods, achieving significant performance gains.",
}
| Open-world Relation Extraction (OpenRE) has recently garnered significant attention. However, existing approaches tend to oversimplify the problem by assuming that all instances of unlabeled data belong to novel classes, thereby limiting the practicality of these methods. We argue that the OpenRE setting should be more aligned with the characteristics of real-world data. Specifically, we propose two key improvements: (a) unlabeled data should encompass known and novel classes, including negative instances; and (b) the set of novel classes should represent long-tail relation types. Furthermore, we observe that popular relations can often be implicitly inferred through specific patterns, while long-tail relations tend to be explicitly expressed. Motivated by these insights, we present a method called KNoRD (Known and Novel Relation Discovery), which effectively classifies explicitly and implicitly expressed relations from known and novel classes within unlabeled data. Experimental evaluations on several Open-world RE benchmarks demonstrate that KNoRD consistently outperforms other existing methods, achieving significant performance gains. | [
"Hogan, William",
"Li, Jiacheng",
"Shang, Jingbo"
] | Open-world Semi-supervised Generalized Relation Discovery Aligned in a Real-world Setting | emnlp-main.880 | 2305.13533 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.881.bib | https://aclanthology.org/2023.emnlp-main.881/ | @inproceedings{zeng-etal-2023-iekg,
title = "{IEKG}: A Commonsense Knowledge Graph for Idiomatic Expressions",
author = "Zeng, Ziheng and
Cheng, Kellen and
Nanniyur, Srihari and
Zhou, Jianing and
Bhat, Suma",
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.881",
doi = "10.18653/v1/2023.emnlp-main.881",
pages = "14243--14264",
abstract = "Idiomatic expression (IE) processing and comprehension have challenged pre-trained language models (PTLMs) because their meanings are non-compositional. Unlike prior works that enable IE comprehension through fine-tuning PTLMs with sentences containing IEs, in this work, we construct IEKG, a commonsense knowledge graph for figurative interpretations of IEs. This extends the established ${ATOMIC}_{20}^{20}$ converting PTLMs into knowledge models (KMs) that encode and infer commonsense knowledge related to IE use. Experiments show that various PTLMs can be converted into KMs with IEKG. We verify the quality of IEKG and the ability of the trained KMs with automatic and human evaluation. Through applications in natural language understanding, we show that a PTLM injected with knowledge from IEKG exhibits improved IE comprehension ability and can generalize to IEs unseen during training.",
}
| Idiomatic expression (IE) processing and comprehension have challenged pre-trained language models (PTLMs) because their meanings are non-compositional. Unlike prior works that enable IE comprehension through fine-tuning PTLMs with sentences containing IEs, in this work, we construct IEKG, a commonsense knowledge graph for figurative interpretations of IEs. This extends the established ${ATOMIC}_{20}^{20}$ converting PTLMs into knowledge models (KMs) that encode and infer commonsense knowledge related to IE use. Experiments show that various PTLMs can be converted into KMs with IEKG. We verify the quality of IEKG and the ability of the trained KMs with automatic and human evaluation. Through applications in natural language understanding, we show that a PTLM injected with knowledge from IEKG exhibits improved IE comprehension ability and can generalize to IEs unseen during training. | [
"Zeng, Ziheng",
"Cheng, Kellen",
"Nanniyur, Srihari",
"Zhou, Jianing",
"Bhat, Suma"
] | IEKG: A Commonsense Knowledge Graph for Idiomatic Expressions | emnlp-main.881 | 2312.06053 | [
"https://github.com/zzeng13/iekg"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.882.bib | https://aclanthology.org/2023.emnlp-main.882/ | @inproceedings{madanagopal-caverlee-2023-bias,
title = "Bias Neutralization in Non-Parallel Texts: A Cyclic Approach with Auxiliary Guidance",
author = "Madanagopal, Karthic and
Caverlee, James",
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.882",
doi = "10.18653/v1/2023.emnlp-main.882",
pages = "14265--14278",
abstract = "Objectivity is a goal for Wikipedia and many news sites, as well as a guiding principle of many large language models. Indeed, several methods have recently been developed for automatic subjective bias neutralization. These methods, however, typically rely on parallel text for training (i.e. a biased sentence coupled with a non-biased sentence), demonstrate poor transfer to new domains, and can lose important bias-independent context. Toward expanding the reach of bias neutralization, we propose in this paper a new approach called FairBalance. Three of its unique features are: i) a cycle consistent adversarial network enables bias neutralization without the need for parallel text; ii) the model design preserves bias-independent content; and iii) through auxiliary guidance, the model highlights sequences of bias-inducing words, yielding strong results in terms of bias neutralization quality. Extensive experiments demonstrate how FairBalance significantly improves subjective bias neutralization compared to other methods.",
}
| Objectivity is a goal for Wikipedia and many news sites, as well as a guiding principle of many large language models. Indeed, several methods have recently been developed for automatic subjective bias neutralization. These methods, however, typically rely on parallel text for training (i.e. a biased sentence coupled with a non-biased sentence), demonstrate poor transfer to new domains, and can lose important bias-independent context. Toward expanding the reach of bias neutralization, we propose in this paper a new approach called FairBalance. Three of its unique features are: i) a cycle consistent adversarial network enables bias neutralization without the need for parallel text; ii) the model design preserves bias-independent content; and iii) through auxiliary guidance, the model highlights sequences of bias-inducing words, yielding strong results in terms of bias neutralization quality. Extensive experiments demonstrate how FairBalance significantly improves subjective bias neutralization compared to other methods. | [
"Madanagopal, Karthic",
"Caverlee, James"
] | Bias Neutralization in Non-Parallel Texts: A Cyclic Approach with Auxiliary Guidance | emnlp-main.882 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.883.bib | https://aclanthology.org/2023.emnlp-main.883/ | @inproceedings{lucas-etal-2023-fighting,
title = "Fighting Fire with Fire: The Dual Role of {LLM}s in Crafting and Detecting Elusive Disinformation",
author = "Lucas, Jason and
Uchendu, Adaku and
Yamashita, Michiharu and
Lee, Jooyoung and
Rohatgi, Shaurya and
Lee, Dongwon",
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.883",
doi = "10.18653/v1/2023.emnlp-main.883",
pages = "14279--14305",
abstract = "Recent ubiquity and disruptive impacts of large language models (LLMs) have raised concerns about their potential to be misused (*.i.e, generating large-scale harmful and misleading content*). To combat this emerging risk of LLMs, we propose a novel {``}***Fighting Fire with Fire***{''} (F3) strategy that harnesses modern LLMs{'} generative and emergent reasoning capabilities to counter human-written and LLM-generated disinformation. First, we leverage GPT-3.5-turbo to synthesize authentic and deceptive LLM-generated content through paraphrase-based and perturbation-based prefix-style prompts, respectively. Second, we apply zero-shot in-context semantic reasoning techniques with cloze-style prompts to discern genuine from deceptive posts and news articles. In our extensive experiments, we observe GPT-3.5-turbo{'}s zero-shot superiority for both in-distribution and out-of-distribution datasets, where GPT-3.5-turbo consistently achieved accuracy at 68-72{\%}, unlike the decline observed in previous customized and fine-tuned disinformation detectors. Our codebase and dataset are available at https://github.com/mickeymst/F3.",
}
| Recent ubiquity and disruptive impacts of large language models (LLMs) have raised concerns about their potential to be misused (*.i.e, generating large-scale harmful and misleading content*). To combat this emerging risk of LLMs, we propose a novel {``}***Fighting Fire with Fire***{''} (F3) strategy that harnesses modern LLMs{'} generative and emergent reasoning capabilities to counter human-written and LLM-generated disinformation. First, we leverage GPT-3.5-turbo to synthesize authentic and deceptive LLM-generated content through paraphrase-based and perturbation-based prefix-style prompts, respectively. Second, we apply zero-shot in-context semantic reasoning techniques with cloze-style prompts to discern genuine from deceptive posts and news articles. In our extensive experiments, we observe GPT-3.5-turbo{'}s zero-shot superiority for both in-distribution and out-of-distribution datasets, where GPT-3.5-turbo consistently achieved accuracy at 68-72{\%}, unlike the decline observed in previous customized and fine-tuned disinformation detectors. Our codebase and dataset are available at https://github.com/mickeymst/F3. | [
"Lucas, Jason",
"Uchendu, Adaku",
"Yamashita, Michiharu",
"Lee, Jooyoung",
"Rohatgi, Shaurya",
"Lee, Dongwon"
] | Fighting Fire with Fire: The Dual Role of LLMs in Crafting and Detecting Elusive Disinformation | emnlp-main.883 | 2310.15515 | [
"https://github.com/mickeymst/f3"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Oral |
|
https://aclanthology.org/2023.emnlp-main.884.bib | https://aclanthology.org/2023.emnlp-main.884/ | @inproceedings{choi-etal-2023-smop,
title = "{SM}o{P}: Towards Efficient and Effective Prompt Tuning with Sparse Mixture-of-Prompts",
author = "Choi, Joon-Young and
Kim, Junho and
Park, Jun-Hyung 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.884",
doi = "10.18653/v1/2023.emnlp-main.884",
pages = "14306--14316",
abstract = "Prompt tuning has emerged as a successful parameter-efficient alternative to the full fine-tuning of language models. However, prior works on prompt tuning often utilize long soft prompts of up to 100 tokens to improve performance, overlooking the inefficiency associated with extended inputs. In this paper, we propose a novel prompt tuning method $SMoP$ ($S$parse $M$ixture-$o$f-$P$rompts) that utilizes short soft prompts for efficient training and inference while maintaining performance gains typically induced from longer soft prompts. To achieve this, $SMoP$ employs a gating mechanism to train multiple short soft prompts specialized in handling different subsets of the data, providing an alternative to relying on a single long soft prompt to cover the entire data. Experimental results demonstrate that $SMoP$ outperforms baseline methods while reducing training and inference costs. We release our code at https://github.com/jyjohnchoi/SMoP.",
}
| Prompt tuning has emerged as a successful parameter-efficient alternative to the full fine-tuning of language models. However, prior works on prompt tuning often utilize long soft prompts of up to 100 tokens to improve performance, overlooking the inefficiency associated with extended inputs. In this paper, we propose a novel prompt tuning method $SMoP$ ($S$parse $M$ixture-$o$f-$P$rompts) that utilizes short soft prompts for efficient training and inference while maintaining performance gains typically induced from longer soft prompts. To achieve this, $SMoP$ employs a gating mechanism to train multiple short soft prompts specialized in handling different subsets of the data, providing an alternative to relying on a single long soft prompt to cover the entire data. Experimental results demonstrate that $SMoP$ outperforms baseline methods while reducing training and inference costs. We release our code at https://github.com/jyjohnchoi/SMoP. | [
"Choi, Joon-Young",
"Kim, Junho",
"Park, Jun-Hyung",
"Mok, Wing-Lam",
"Lee, SangKeun"
] | SMoP: Towards Efficient and Effective Prompt Tuning with Sparse Mixture-of-Prompts | emnlp-main.884 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.885.bib | https://aclanthology.org/2023.emnlp-main.885/ | @inproceedings{jiang-etal-2023-brainteaser,
title = "{BRAINTEASER}: Lateral Thinking Puzzles for Large Language Models",
author = "Jiang, Yifan and
Ilievski, Filip and
Ma, Kaixin and
Sourati, Zhivar",
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.885",
doi = "10.18653/v1/2023.emnlp-main.885",
pages = "14317--14332",
abstract = "The success of language models has inspired the NLP community to attend to tasks that require implicit and complex reasoning, relying on human-like commonsense mechanisms. While such vertical thinking tasks have been relatively popular, lateral thinking puzzles have received little attention. To bridge this gap, we devise BrainTeaser: a multiple-choice Question Answering task designed to test the model{'}s ability to exhibit lateral thinking and defy default commonsense associations. We design a three-step procedure for creating the first lateral thinking benchmark, consisting of data collection, distractor generation, and generation of adversarial examples, leading to 1,100 puzzles with high-quality annotations. To assess the consistency of lateral reasoning by models, we enrich BrainTeaser based on a semantic and contextual reconstruction of its questions. Our experiments with state-of-the-art instruction- and commonsense language models reveal a significant gap between human and model performance, which is further widened when consistency across adversarial formats is considered. We make all of our code and data available to stimulate work on developing and evaluating lateral thinking models.",
}
| The success of language models has inspired the NLP community to attend to tasks that require implicit and complex reasoning, relying on human-like commonsense mechanisms. While such vertical thinking tasks have been relatively popular, lateral thinking puzzles have received little attention. To bridge this gap, we devise BrainTeaser: a multiple-choice Question Answering task designed to test the model{'}s ability to exhibit lateral thinking and defy default commonsense associations. We design a three-step procedure for creating the first lateral thinking benchmark, consisting of data collection, distractor generation, and generation of adversarial examples, leading to 1,100 puzzles with high-quality annotations. To assess the consistency of lateral reasoning by models, we enrich BrainTeaser based on a semantic and contextual reconstruction of its questions. Our experiments with state-of-the-art instruction- and commonsense language models reveal a significant gap between human and model performance, which is further widened when consistency across adversarial formats is considered. We make all of our code and data available to stimulate work on developing and evaluating lateral thinking models. | [
"Jiang, Yifan",
"Ilievski, Filip",
"Ma, Kaixin",
"Sourati, Zhivar"
] | BRAINTEASER: Lateral Thinking Puzzles for Large Language Models | emnlp-main.885 | 2310.05057 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.886.bib | https://aclanthology.org/2023.emnlp-main.886/ | @inproceedings{tang-etal-2023-lemons,
title = "When are Lemons Purple? The Concept Association Bias of Vision-Language Models",
author = "Tang, Yingtian and
Yamada, Yutaro and
Zhang, Yoyo and
Yildirim, Ilker",
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.886",
doi = "10.18653/v1/2023.emnlp-main.886",
pages = "14333--14348",
abstract = "Large-scale vision-language models such as CLIP have shown impressive performance on zero-shot image classification and image-to-text retrieval. However, such performance does not realize in tasks that require a finer-grained correspondence between vision and language, such as Visual Question Answering (VQA). We investigate why this is the case, and report an interesting phenomenon of vision-language models, which we call the Concept Association Bias (CAB), as a potential cause of the difficulty of applying these models to VQA and similar tasks. We find that models with CAB tend to treat input as a bag of concepts and attempt to fill in the other missing concept crossmodally, leading to an unexpected zero-shot prediction. We demonstrate CAB by showing that CLIP{'}s zero-shot classification performance greatly suffers when there is a strong concept association between an object (e.g. eggplant) and an attribute (e.g. color purple). We also show that the strength of CAB predicts the performance on VQA. We observe that CAB is prevalent in vision-language models trained with contrastive losses, even when autoregressive losses are jointly employed. However, a model that solely relies on autoregressive loss seems to exhibit minimal or no signs of CAB.",
}
| Large-scale vision-language models such as CLIP have shown impressive performance on zero-shot image classification and image-to-text retrieval. However, such performance does not realize in tasks that require a finer-grained correspondence between vision and language, such as Visual Question Answering (VQA). We investigate why this is the case, and report an interesting phenomenon of vision-language models, which we call the Concept Association Bias (CAB), as a potential cause of the difficulty of applying these models to VQA and similar tasks. We find that models with CAB tend to treat input as a bag of concepts and attempt to fill in the other missing concept crossmodally, leading to an unexpected zero-shot prediction. We demonstrate CAB by showing that CLIP{'}s zero-shot classification performance greatly suffers when there is a strong concept association between an object (e.g. eggplant) and an attribute (e.g. color purple). We also show that the strength of CAB predicts the performance on VQA. We observe that CAB is prevalent in vision-language models trained with contrastive losses, even when autoregressive losses are jointly employed. However, a model that solely relies on autoregressive loss seems to exhibit minimal or no signs of CAB. | [
"Tang, Yingtian",
"Yamada, Yutaro",
"Zhang, Yoyo",
"Yildirim, Ilker"
] | When are Lemons Purple? The Concept Association Bias of Vision-Language Models | emnlp-main.886 | 2212.12043 | [
""
] | https://huggingface.co/papers/2212.12043 | 1 | 0 | 0 | 3 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.887.bib | https://aclanthology.org/2023.emnlp-main.887/ | @inproceedings{giulianelli-etal-2023-comes,
title = "What Comes Next? Evaluating Uncertainty in Neural Text Generators Against Human Production Variability",
author = "Giulianelli, Mario and
Baan, Joris and
Aziz, Wilker and
Fern{\'a}ndez, Raquel and
Plank, Barbara",
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.887",
doi = "10.18653/v1/2023.emnlp-main.887",
pages = "14349--14371",
abstract = "In Natural Language Generation (NLG) tasks, for any input, multiple communicative goals are plausible, and any goal can be put into words, or produced, in multiple ways. We characterise the extent to which human production varies lexically, syntactically, and semantically across four NLG tasks, connecting human production variability to aleatoric or data uncertainty. We then inspect the space of output strings shaped by a generation system{'}s predicted probability distribution and decoding algorithm to probe its uncertainty. For each test input, we measure the generator{'}s calibration to human production variability. Following this instance-level approach, we analyse NLG models and decoding strategies, demonstrating that probing a generator with multiple samples and, when possible, multiple references, provides the level of detail necessary to gain understanding of a model{'}s representation of uncertainty.",
}
| In Natural Language Generation (NLG) tasks, for any input, multiple communicative goals are plausible, and any goal can be put into words, or produced, in multiple ways. We characterise the extent to which human production varies lexically, syntactically, and semantically across four NLG tasks, connecting human production variability to aleatoric or data uncertainty. We then inspect the space of output strings shaped by a generation system{'}s predicted probability distribution and decoding algorithm to probe its uncertainty. For each test input, we measure the generator{'}s calibration to human production variability. Following this instance-level approach, we analyse NLG models and decoding strategies, demonstrating that probing a generator with multiple samples and, when possible, multiple references, provides the level of detail necessary to gain understanding of a model{'}s representation of uncertainty. | [
"Giulianelli, Mario",
"Baan, Joris",
"Aziz, Wilker",
"Fern{\\'a}ndez, Raquel",
"Plank, Barbara"
] | What Comes Next? Evaluating Uncertainty in Neural Text Generators Against Human Production Variability | emnlp-main.887 | 2305.11707 | [
"https://github.com/dmg-illc/nlg-uncertainty-probes"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Oral |
|
https://aclanthology.org/2023.emnlp-main.888.bib | https://aclanthology.org/2023.emnlp-main.888/ | @inproceedings{zhang-etal-2023-text,
title = "Text Representation Distillation via Information Bottleneck Principle",
author = "Zhang, Yanzhao and
Long, Dingkun and
Li, Zehan and
Xie, Pengjun",
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.888",
doi = "10.18653/v1/2023.emnlp-main.888",
pages = "14372--14383",
abstract = "Pre-trained language models (PLMs) have recently shown great success in text representation field. However, the high computational cost and high-dimensional representation of PLMs pose significant challenges for practical applications. To make models more accessible, an effective method is to distill large models into smaller representation models. In order to relieve the issue of performance degradation after distillation, we propose a novel Knowledge Distillation method called \textbf{IBKD}. This approach is motivated by the Information Bottleneck principle and aims to maximize the mutual information between the final representation of the teacher and student model, while simultaneously reducing the mutual information between the student model{'}s representation and the input data. This enables the student model to preserve important learned information while avoiding unnecessary information, thus reducing the risk of over-fitting. Empirical studies on two main downstream applications of text representation (Semantic Textual Similarity and Dense Retrieval tasks) demonstrate the effectiveness of our proposed approach.",
}
| Pre-trained language models (PLMs) have recently shown great success in text representation field. However, the high computational cost and high-dimensional representation of PLMs pose significant challenges for practical applications. To make models more accessible, an effective method is to distill large models into smaller representation models. In order to relieve the issue of performance degradation after distillation, we propose a novel Knowledge Distillation method called \textbf{IBKD}. This approach is motivated by the Information Bottleneck principle and aims to maximize the mutual information between the final representation of the teacher and student model, while simultaneously reducing the mutual information between the student model{'}s representation and the input data. This enables the student model to preserve important learned information while avoiding unnecessary information, thus reducing the risk of over-fitting. Empirical studies on two main downstream applications of text representation (Semantic Textual Similarity and Dense Retrieval tasks) demonstrate the effectiveness of our proposed approach. | [
"Zhang, Yanzhao",
"Long, Dingkun",
"Li, Zehan",
"Xie, Pengjun"
] | Text Representation Distillation via Information Bottleneck Principle | emnlp-main.888 | 2311.05472 | [
"https://github.com/alibaba-nlp/ibkd"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.889.bib | https://aclanthology.org/2023.emnlp-main.889/ | @inproceedings{liang-etal-2023-gpt,
title = "Let {GPT} be a Math Tutor: Teaching Math Word Problem Solvers with Customized Exercise Generation",
author = "Liang, Zhenwen and
Yu, Wenhao and
Rajpurohit, Tanmay and
Clark, Peter and
Zhang, Xiangliang and
Kalyan, Ashwin",
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.889",
doi = "10.18653/v1/2023.emnlp-main.889",
pages = "14384--14396",
abstract = "In this paper, we present a novel approach for distilling math word problem solving capabilities from large language models (LLMs) into smaller, more efficient student models. Our approach is designed to consider the student model{'}s weaknesses and foster a tailored learning experience by generating targeted exercises aligned with educational science principles, such as knowledge tracing and personalized learning. Concretely, we let GPT-3 be a math tutor and run two steps iteratively: 1) assessing the student model{'}s current learning status on a GPT-generated exercise book, and 2) improving the student model by training it with tailored exercise samples generated by GPT-3. Experimental results reveal that our approach outperforms LLMs (e.g., GPT-3 and PaLM) in accuracy across three distinct benchmarks while employing significantly fewer parameters. Furthermore, we provide a comprehensive analysis of the various components within our methodology to substantiate their efficacy.",
}
| In this paper, we present a novel approach for distilling math word problem solving capabilities from large language models (LLMs) into smaller, more efficient student models. Our approach is designed to consider the student model{'}s weaknesses and foster a tailored learning experience by generating targeted exercises aligned with educational science principles, such as knowledge tracing and personalized learning. Concretely, we let GPT-3 be a math tutor and run two steps iteratively: 1) assessing the student model{'}s current learning status on a GPT-generated exercise book, and 2) improving the student model by training it with tailored exercise samples generated by GPT-3. Experimental results reveal that our approach outperforms LLMs (e.g., GPT-3 and PaLM) in accuracy across three distinct benchmarks while employing significantly fewer parameters. Furthermore, we provide a comprehensive analysis of the various components within our methodology to substantiate their efficacy. | [
"Liang, Zhenwen",
"Yu, Wenhao",
"Rajpurohit, Tanmay",
"Clark, Peter",
"Zhang, Xiangliang",
"Kalyan, Ashwin"
] | Let GPT be a Math Tutor: Teaching Math Word Problem Solvers with Customized Exercise Generation | emnlp-main.889 | 2305.14386 | [
""
] | https://huggingface.co/papers/2305.14386 | 1 | 0 | 0 | 6 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.890.bib | https://aclanthology.org/2023.emnlp-main.890/ | @inproceedings{kim-etal-2023-fantom,
title = "{FANT}o{M}: A Benchmark for Stress-testing Machine Theory of Mind in Interactions",
author = "Kim, Hyunwoo and
Sclar, Melanie and
Zhou, Xuhui and
Bras, Ronan and
Kim, Gunhee and
Choi, Yejin and
Sap, Maarten",
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.890",
doi = "10.18653/v1/2023.emnlp-main.890",
pages = "14397--14413",
abstract = "Theory of mind (ToM) evaluations currently focus on testing models using passive narratives that inherently lack interactivity. We introduce FANToM, a new benchmark designed to stress-test ToM within information-asymmetric conversational contexts via question answering. Our benchmark draws upon important theoretical requisites from psychology and necessary empirical considerations when evaluating large language models (LLMs). In particular, we formulate multiple types of questions that demand the same underlying reasoning to identify illusory or false sense of ToM capabilities in LLMs. We show that FANToM is challenging for state-of-the-art LLMs, which perform significantly worse than humans even with chain-of-thought reasoning or fine-tuning.",
}
| Theory of mind (ToM) evaluations currently focus on testing models using passive narratives that inherently lack interactivity. We introduce FANToM, a new benchmark designed to stress-test ToM within information-asymmetric conversational contexts via question answering. Our benchmark draws upon important theoretical requisites from psychology and necessary empirical considerations when evaluating large language models (LLMs). In particular, we formulate multiple types of questions that demand the same underlying reasoning to identify illusory or false sense of ToM capabilities in LLMs. We show that FANToM is challenging for state-of-the-art LLMs, which perform significantly worse than humans even with chain-of-thought reasoning or fine-tuning. | [
"Kim, Hyunwoo",
"Sclar, Melanie",
"Zhou, Xuhui",
"Bras, Ronan",
"Kim, Gunhee",
"Choi, Yejin",
"Sap, Maarten"
] | FANToM: A Benchmark for Stress-testing Machine Theory of Mind in Interactions | emnlp-main.890 | 2310.15421 | [
""
] | https://huggingface.co/papers/2310.15421 | 2 | 0 | 0 | 7 | [] | [] | [] | 1 | Oral |
https://aclanthology.org/2023.emnlp-main.891.bib | https://aclanthology.org/2023.emnlp-main.891/ | @inproceedings{liu-etal-2023-exploring-boundaries,
title = "Exploring the Boundaries of {GPT}-4 in Radiology",
author = "Liu, Qianchu and
Hyland, Stephanie and
Bannur, Shruthi and
Bouzid, Kenza and
Castro, Daniel and
Wetscherek, Maria and
Tinn, Robert and
Sharma, Harshita and
P{\'e}rez-Garc{\'\i}a, Fernando and
Schwaighofer, Anton and
Rajpurkar, Pranav and
Khanna, Sameer and
Poon, Hoifung and
Usuyama, Naoto and
Thieme, Anja and
Nori, Aditya and
Lungren, Matthew and
Oktay, Ozan and
Alvarez-Valle, Javier",
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.891",
doi = "10.18653/v1/2023.emnlp-main.891",
pages = "14414--14445",
abstract = "The recent success of general-domain large language models (LLMs) has significantly changed the natural language processing paradigm towards a unified foundation model across domains and applications. In this paper, we focus on assessing the performance of GPT-4, the most capable LLM so far, on the text-based applications for radiology reports, comparing against state-of-the-art (SOTA) radiology-specific models. Exploring various prompting strategies, we evaluated GPT-4 on a diverse range of common radiology tasks and we found GPT-4 either outperforms or is on par with current SOTA radiology models. With zero-shot prompting, GPT-4 already obtains substantial gains ($\approx$ 10{\%} absolute improvement) over radiology models in temporal sentence similarity classification (accuracy) and natural language inference ($F_1$). For tasks that require learning dataset-specific style or schema (e.g. findings summarisation), GPT-4 improves with example-based prompting and matches supervised SOTA. Our extensive error analysis with a board-certified radiologist shows GPT-4 has a sufficient level of radiology knowledge with only occasional errors in complex context that require nuanced domain knowledge. For findings summarisation, GPT-4 outputs are found to be overall comparable with existing manually-written impressions.",
}
| The recent success of general-domain large language models (LLMs) has significantly changed the natural language processing paradigm towards a unified foundation model across domains and applications. In this paper, we focus on assessing the performance of GPT-4, the most capable LLM so far, on the text-based applications for radiology reports, comparing against state-of-the-art (SOTA) radiology-specific models. Exploring various prompting strategies, we evaluated GPT-4 on a diverse range of common radiology tasks and we found GPT-4 either outperforms or is on par with current SOTA radiology models. With zero-shot prompting, GPT-4 already obtains substantial gains ($\approx$ 10{\%} absolute improvement) over radiology models in temporal sentence similarity classification (accuracy) and natural language inference ($F_1$). For tasks that require learning dataset-specific style or schema (e.g. findings summarisation), GPT-4 improves with example-based prompting and matches supervised SOTA. Our extensive error analysis with a board-certified radiologist shows GPT-4 has a sufficient level of radiology knowledge with only occasional errors in complex context that require nuanced domain knowledge. For findings summarisation, GPT-4 outputs are found to be overall comparable with existing manually-written impressions. | [
"Liu, Qianchu",
"Hyl",
", Stephanie",
"Bannur, Shruthi",
"Bouzid, Kenza",
"Castro, Daniel",
"Wetscherek, Maria",
"Tinn, Robert",
"Sharma, Harshita",
"P{\\'e}rez-Garc{\\'\\i}a, Fern",
"o",
"Schwaighofer, Anton",
"Rajpurkar, Pranav",
"Khanna, Sameer",
"Poon, Hoifung",
"Usuyama, Naoto",
"Thieme, Anja",
"Nori, Aditya",
"Lungren, Matthew",
"Oktay, Ozan",
"Alvarez-Valle, Javier"
] | Exploring the Boundaries of GPT-4 in Radiology | emnlp-main.891 | 2310.14573 | [
""
] | https://huggingface.co/papers/2310.14573 | 9 | 7 | 2 | 19 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.892.bib | https://aclanthology.org/2023.emnlp-main.892/ | @inproceedings{jeon-etal-2023-frustratingly,
title = "A Frustratingly Easy Post-Training Quantization Scheme for {LLM}s",
author = "Jeon, Yongkweon and
Lee, Chungman and
Park, Kyungphil and
Kim, Ho-young",
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.892",
doi = "10.18653/v1/2023.emnlp-main.892",
pages = "14446--14461",
abstract = "Efficient inference has become crucial for hyper-scale AI models, including large language models, as their parameter count continues to increase for enhanced performance. This necessity holds true regardless of the computing environment, whether it be mobile devices or cloud servers. Quantization emerges as a solution to alleviate the computational burden during inference. By representing models with a reduced bit-width, quantization minimizes the frequency of DRAM access while fully exploiting the parallelism of operations through a dense matrix format. Consequently, quantized models achieve low end-to-end latency and optimize resource utilization by addressing both memory and computing bottlenecks. In this paper, we propose a straightforward post-training quantization scheme, called Z-Fold, that fully utilizes the feature of the Transformer structure widely employed in large language models.",
}
| Efficient inference has become crucial for hyper-scale AI models, including large language models, as their parameter count continues to increase for enhanced performance. This necessity holds true regardless of the computing environment, whether it be mobile devices or cloud servers. Quantization emerges as a solution to alleviate the computational burden during inference. By representing models with a reduced bit-width, quantization minimizes the frequency of DRAM access while fully exploiting the parallelism of operations through a dense matrix format. Consequently, quantized models achieve low end-to-end latency and optimize resource utilization by addressing both memory and computing bottlenecks. In this paper, we propose a straightforward post-training quantization scheme, called Z-Fold, that fully utilizes the feature of the Transformer structure widely employed in large language models. | [
"Jeon, Yongkweon",
"Lee, Chungman",
"Park, Kyungphil",
"Kim, Ho-young"
] | A Frustratingly Easy Post-Training Quantization Scheme for LLMs | emnlp-main.892 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.893.bib | https://aclanthology.org/2023.emnlp-main.893/ | @inproceedings{kartchner-etal-2023-comprehensive,
title = "A Comprehensive Evaluation of Biomedical Entity Linking Models",
author = "Kartchner, David and
Deng, Jennifer and
Lohiya, Shubham and
Kopparthi, Tejasri and
Bathala, Prasanth and
Domingo-Fern{\'a}ndez, Daniel and
Mitchell, Cassie",
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.893",
doi = "10.18653/v1/2023.emnlp-main.893",
pages = "14462--14478",
abstract = "Biomedical entity linking (BioEL) is the process of connecting entities referenced in documents to entries in biomedical databases such as the Unified Medical Language System (UMLS) or Medical Subject Headings (MeSH). The study objective was to comprehensively evaluate nine recent state-of-the-art biomedical entity linking models under a unified framework. We compare these models along axes of (1) accuracy, (2) speed, (3) ease of use, (4) generalization, and (5) adaptability to new ontologies and datasets. We additionally quantify the impact of various preprocessing choices such as abbreviation detection. Systematic evaluation reveals several notable gaps in current methods. In particular, current methods struggle to correctly link genes and proteins and often have difficulty effectively incorporating context into linking decisions. To expedite future development and baseline testing, we release our unified evaluation framework and all included models on GitHub at https://github.com/davidkartchner/biomedical-entity-linking",
}
| Biomedical entity linking (BioEL) is the process of connecting entities referenced in documents to entries in biomedical databases such as the Unified Medical Language System (UMLS) or Medical Subject Headings (MeSH). The study objective was to comprehensively evaluate nine recent state-of-the-art biomedical entity linking models under a unified framework. We compare these models along axes of (1) accuracy, (2) speed, (3) ease of use, (4) generalization, and (5) adaptability to new ontologies and datasets. We additionally quantify the impact of various preprocessing choices such as abbreviation detection. Systematic evaluation reveals several notable gaps in current methods. In particular, current methods struggle to correctly link genes and proteins and often have difficulty effectively incorporating context into linking decisions. To expedite future development and baseline testing, we release our unified evaluation framework and all included models on GitHub at https://github.com/davidkartchner/biomedical-entity-linking | [
"Kartchner, David",
"Deng, Jennifer",
"Lohiya, Shubham",
"Kopparthi, Tejasri",
"Bathala, Prasanth",
"Domingo-Fern{\\'a}ndez, Daniel",
"Mitchell, Cassie"
] | A Comprehensive Evaluation of Biomedical Entity Linking Models | emnlp-main.893 | null | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.894.bib | https://aclanthology.org/2023.emnlp-main.894/ | @inproceedings{purkayastha-etal-2023-exploring,
title = "Exploring Jiu-Jitsu Argumentation for Writing Peer Review Rebuttals",
author = "Purkayastha, Sukannya and
Lauscher, Anne and
Gurevych, Iryna",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.emnlp-main.894",
doi = "10.18653/v1/2023.emnlp-main.894",
pages = "14479--14495",
abstract = "In many domains of argumentation, people{'}s arguments are driven by so-called attitude roots, i.e., underlying beliefs and world views, and their corresponding attitude themes. Given the strength of these latent drivers of arguments, recent work in psychology suggests that instead of directly countering surface-level reasoning (e.g., falsifying the premises), one should follow an argumentation style inspired by the Jiu-Jitsu {``}soft{''} combat system: first, identify an arguer{'}s attitude roots and themes, and then choose a prototypical rebuttal that is aligned with those drivers instead of trying to invalidate those. In this work, we are the first to explore Jiu-Jitsu argumentation for peer reviews by proposing the novel task of attitude and theme-guided rebuttal generation. To this end, we enrich an existing dataset for discourse structure in peer reviews with attitude roots, attitude themes, and canonical rebuttals. To facilitate this process, we recast established annotation concepts from the domain of peer reviews (e.g., aspects a review sentence is relating to) and train domain-specific models. We then propose strong rebuttal generation strategies, which we benchmark on our novel dataset for the task of end-to-end attitude and theme-guided rebuttal generation and two subtasks.",
}
| In many domains of argumentation, people{'}s arguments are driven by so-called attitude roots, i.e., underlying beliefs and world views, and their corresponding attitude themes. Given the strength of these latent drivers of arguments, recent work in psychology suggests that instead of directly countering surface-level reasoning (e.g., falsifying the premises), one should follow an argumentation style inspired by the Jiu-Jitsu {``}soft{''} combat system: first, identify an arguer{'}s attitude roots and themes, and then choose a prototypical rebuttal that is aligned with those drivers instead of trying to invalidate those. In this work, we are the first to explore Jiu-Jitsu argumentation for peer reviews by proposing the novel task of attitude and theme-guided rebuttal generation. To this end, we enrich an existing dataset for discourse structure in peer reviews with attitude roots, attitude themes, and canonical rebuttals. To facilitate this process, we recast established annotation concepts from the domain of peer reviews (e.g., aspects a review sentence is relating to) and train domain-specific models. We then propose strong rebuttal generation strategies, which we benchmark on our novel dataset for the task of end-to-end attitude and theme-guided rebuttal generation and two subtasks. | [
"Purkayastha, Sukannya",
"Lauscher, Anne",
"Gurevych, Iryna"
] | Exploring Jiu-Jitsu Argumentation for Writing Peer Review Rebuttals | emnlp-main.894 | 2311.03998 | [
"https://github.com/ukplab/emnlp2023_jiu_jitsu_argumentation_for_rebuttals"
] | https://huggingface.co/papers/2311.03998 | 1 | 0 | 0 | 3 | [] | [] | [] | 1 | Poster |
https://aclanthology.org/2023.emnlp-main.895.bib | https://aclanthology.org/2023.emnlp-main.895/ | @inproceedings{agarwal-etal-2023-limit,
title = "{LIMIT}: Language Identification, Misidentification, and Translation using Hierarchical Models in 350+ Languages",
author = "Agarwal, Milind and
Alam, Md Mahfuz Ibn and
Anastasopoulos, Antonios",
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.895",
doi = "10.18653/v1/2023.emnlp-main.895",
pages = "14496--14519",
abstract = "Knowing the language of an input text/audio is a necessary first step for using almost every NLP tool such as taggers, parsers, or translation systems. Language identification is a well-studied problem, sometimes even considered solved; in reality, due to lack of data and computational challenges, current systems cannot accurately identify most of the world{'}s 7000 languages. To tackle this bottleneck, we first compile a corpus, MCS-350, of 50K multilingual and parallel children{'}s stories in 350+ languages. MCS-350 can serve as a benchmark for language identification of short texts and for 1400+ new translation directions in low-resource Indian and African languages. Second, we propose a novel misprediction-resolution hierarchical model, LIMIT, for language identification that reduces error by 55{\%} (from 0.71 to 0.32) on our compiled children{'}s stories dataset and by 40{\%} (from 0.23 to 0.14) on the FLORES-200 benchmark. Our method can expand language identification coverage into low-resource languages by relying solely on systemic misprediction patterns, bypassing the need to retrain large models from scratch.",
}
| Knowing the language of an input text/audio is a necessary first step for using almost every NLP tool such as taggers, parsers, or translation systems. Language identification is a well-studied problem, sometimes even considered solved; in reality, due to lack of data and computational challenges, current systems cannot accurately identify most of the world{'}s 7000 languages. To tackle this bottleneck, we first compile a corpus, MCS-350, of 50K multilingual and parallel children{'}s stories in 350+ languages. MCS-350 can serve as a benchmark for language identification of short texts and for 1400+ new translation directions in low-resource Indian and African languages. Second, we propose a novel misprediction-resolution hierarchical model, LIMIT, for language identification that reduces error by 55{\%} (from 0.71 to 0.32) on our compiled children{'}s stories dataset and by 40{\%} (from 0.23 to 0.14) on the FLORES-200 benchmark. Our method can expand language identification coverage into low-resource languages by relying solely on systemic misprediction patterns, bypassing the need to retrain large models from scratch. | [
"Agarwal, Milind",
"Alam, Md Mahfuz Ibn",
"Anastasopoulos, Antonios"
] | LIMIT: Language Identification, Misidentification, and Translation using Hierarchical Models in 350+ Languages | emnlp-main.895 | 2305.14263 | [
"https://github.com/magarw/limit"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.896.bib | https://aclanthology.org/2023.emnlp-main.896/ | @inproceedings{xiao-etal-2023-freeal,
title = "{F}ree{AL}: Towards Human-Free Active Learning in the Era of Large Language Models",
author = "Xiao, Ruixuan and
Dong, Yiwen and
Zhao, Junbo and
Wu, Runze and
Lin, Minmin and
Chen, Gang and
Wang, Haobo",
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.896",
doi = "10.18653/v1/2023.emnlp-main.896",
pages = "14520--14535",
abstract = "Collecting high-quality labeled data for model training is notoriously time-consuming and labor-intensive for various NLP tasks. While copious solutions, such as active learning for small language models (SLMs) and prevalent in-context learning in the era of large language models (LLMs), have been proposed and alleviate the labeling burden to some extent, their performances are still subject to human intervention. It is still underexplored how to reduce the annotation cost in the LLMs era. To bridge this, we revolutionize traditional active learning and propose an innovative collaborative learning framework FreeAL to interactively distill and filter the task-specific knowledge from LLMs. During collaborative training, an LLM serves as an active annotator inculcating its coarse-grained knowledge, while a downstream SLM is incurred as a student to filter out high-quality in-context samples to feedback LLM for the subsequent label refinery. Extensive experiments on eight benchmark datasets demonstrate that FreeAL largely enhances the zero-shot performances for both SLM and LLM without any human supervision.",
}
| Collecting high-quality labeled data for model training is notoriously time-consuming and labor-intensive for various NLP tasks. While copious solutions, such as active learning for small language models (SLMs) and prevalent in-context learning in the era of large language models (LLMs), have been proposed and alleviate the labeling burden to some extent, their performances are still subject to human intervention. It is still underexplored how to reduce the annotation cost in the LLMs era. To bridge this, we revolutionize traditional active learning and propose an innovative collaborative learning framework FreeAL to interactively distill and filter the task-specific knowledge from LLMs. During collaborative training, an LLM serves as an active annotator inculcating its coarse-grained knowledge, while a downstream SLM is incurred as a student to filter out high-quality in-context samples to feedback LLM for the subsequent label refinery. Extensive experiments on eight benchmark datasets demonstrate that FreeAL largely enhances the zero-shot performances for both SLM and LLM without any human supervision. | [
"Xiao, Ruixuan",
"Dong, Yiwen",
"Zhao, Junbo",
"Wu, Runze",
"Lin, Minmin",
"Chen, Gang",
"Wang, Haobo"
] | FreeAL: Towards Human-Free Active Learning in the Era of Large Language Models | emnlp-main.896 | 2311.15614 | [
"https://github.com/justherozen/freeal"
] | https://huggingface.co/papers/2311.15614 | 0 | 0 | 0 | 7 | [] | [] | [] | 1 | Oral |
https://aclanthology.org/2023.emnlp-main.897.bib | https://aclanthology.org/2023.emnlp-main.897/ | @inproceedings{cao-etal-2023-api,
title = "{API}-Assisted Code Generation for Question Answering on Varied Table Structures",
author = "Cao, Yihan and
Chen, Shuyi and
Liu, Ryan and
Wang, Zhiruo and
Fried, Daniel",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.emnlp-main.897",
doi = "10.18653/v1/2023.emnlp-main.897",
pages = "14536--14548",
abstract = "A persistent challenge to table question answering (TableQA) by generating executable programs has been adapting to varied table structures, typically requiring domain-specific logical forms. In response, this paper introduces a unified TableQA framework that: (1) provides a unified representation for structured tables as multi-index Pandas data frames, (2) uses Python as a powerful querying language, and (3) uses few-shot prompting to translate NL questions into Python programs, which are executable on Pandas data frames. Furthermore, to answer complex relational questions with extended program functionality and external knowledge, our framework allows customized APIs that Python programs can call. We experiment with four TableQA datasets that involve tables of different structures {---} relational, multi-table, and hierarchical matrix shapes {---} and achieve prominent improvements over past state-of-the-art systems. In ablation studies, we (1) show benefits from our multi-index representation and APIs over baselines that use only an LLM, and (2) demonstrate that our approach is modular and can incorporate additional APIs.",
}
| A persistent challenge to table question answering (TableQA) by generating executable programs has been adapting to varied table structures, typically requiring domain-specific logical forms. In response, this paper introduces a unified TableQA framework that: (1) provides a unified representation for structured tables as multi-index Pandas data frames, (2) uses Python as a powerful querying language, and (3) uses few-shot prompting to translate NL questions into Python programs, which are executable on Pandas data frames. Furthermore, to answer complex relational questions with extended program functionality and external knowledge, our framework allows customized APIs that Python programs can call. We experiment with four TableQA datasets that involve tables of different structures {---} relational, multi-table, and hierarchical matrix shapes {---} and achieve prominent improvements over past state-of-the-art systems. In ablation studies, we (1) show benefits from our multi-index representation and APIs over baselines that use only an LLM, and (2) demonstrate that our approach is modular and can incorporate additional APIs. | [
"Cao, Yihan",
"Chen, Shuyi",
"Liu, Ryan",
"Wang, Zhiruo",
"Fried, Daniel"
] | API-Assisted Code Generation for Question Answering on Varied Table Structures | emnlp-main.897 | 2310.14687 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.898.bib | https://aclanthology.org/2023.emnlp-main.898/ | @inproceedings{zhou-etal-2023-data,
title = "Data Factors for Better Compositional Generalization",
author = "Zhou, Xiang and
Jiang, Yichen 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.898",
doi = "10.18653/v1/2023.emnlp-main.898",
pages = "14549--14566",
abstract = "Recent diagnostic datasets on compositional generalization, such as SCAN (Lake and Baroni, 2018) and COGS (Kim and Linzen, 2020), expose severe problems in models trained from scratch on these datasets. However, in contrast to this poor performance, state-of-the-art models trained on larger and more general datasets show better generalization ability. In this work, to reconcile this inconsistency, we conduct an empirical analysis by training Transformer models on a variety of training sets with different data factors, including dataset scale, pattern complexity, example difficulty, etc. First, we show that increased dataset complexity can lead to better generalization behavior on multiple different generalization challenges. To further understand this improvement, we show two axes of the benefit from more complex datasets: they provide more diverse examples so compositional understanding becomes more effective, and they also prevent ungeneralizable memorization of the examples due to reduced example repetition frequency. Finally, we explore how training examples of different difficulty levels influence generalization differently. On synthetic datasets, simple examples invoke stronger compositionality than hard examples do. On larger-scale real language datasets, while hard examples become more important potentially to ensure decent data coverage, a balanced mixture of simple and hard examples manages to induce the strongest generalizability.",
}
| Recent diagnostic datasets on compositional generalization, such as SCAN (Lake and Baroni, 2018) and COGS (Kim and Linzen, 2020), expose severe problems in models trained from scratch on these datasets. However, in contrast to this poor performance, state-of-the-art models trained on larger and more general datasets show better generalization ability. In this work, to reconcile this inconsistency, we conduct an empirical analysis by training Transformer models on a variety of training sets with different data factors, including dataset scale, pattern complexity, example difficulty, etc. First, we show that increased dataset complexity can lead to better generalization behavior on multiple different generalization challenges. To further understand this improvement, we show two axes of the benefit from more complex datasets: they provide more diverse examples so compositional understanding becomes more effective, and they also prevent ungeneralizable memorization of the examples due to reduced example repetition frequency. Finally, we explore how training examples of different difficulty levels influence generalization differently. On synthetic datasets, simple examples invoke stronger compositionality than hard examples do. On larger-scale real language datasets, while hard examples become more important potentially to ensure decent data coverage, a balanced mixture of simple and hard examples manages to induce the strongest generalizability. | [
"Zhou, Xiang",
"Jiang, Yichen",
"Bansal, Mohit"
] | Data Factors for Better Compositional Generalization | emnlp-main.898 | 2311.04420 | [
"https://github.com/owenzx/data4comp"
] | https://huggingface.co/papers/2311.04420 | 0 | 0 | 0 | 3 | [] | [] | [] | 1 | Oral |
https://aclanthology.org/2023.emnlp-main.899.bib | https://aclanthology.org/2023.emnlp-main.899/ | @inproceedings{cui-etal-2023-chatedit,
title = "{C}hat{E}dit: Towards Multi-turn Interactive Facial Image Editing via Dialogue",
author = "Cui, Xing and
Li, Zekun and
Li, Pei and
Hu, Yibo and
Shi, Hailin and
Cao, Chunshui and
He, Zhaofeng",
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.899",
doi = "10.18653/v1/2023.emnlp-main.899",
pages = "14567--14583",
abstract = "This paper explores interactive facial image editing through dialogue and presents the ChatEdit benchmark dataset for evaluating image editing and conversation abilities in this context. ChatEdit is constructed from the CelebA-HQ dataset, incorporating annotated multi-turn dialogues corresponding to user editing requests on the images. The dataset is challenging, as it requires the system to dynamically track and edit images based on user requests, while generating appropriate natural language responses. To address these challenges, we propose a framework comprising a dialogue module for tracking user requests as well as generating responses, and an image editing module for editing images accordingly. Unlike previous approaches, our framework directly tracks the user request of the current turn from the entire dialogue history and edits the initial image instead of manipulating the output from the previous turn, mitigating error accumulation and attribute forgetting issues. Extensive experiments on the ChatEdit dataset demonstrate the superiority of our framework over previous methods and also improvement rooms, encouraging future research. We will release the code and data publicly to facilitate advancements in complex interactive facial image editing.",
}
| This paper explores interactive facial image editing through dialogue and presents the ChatEdit benchmark dataset for evaluating image editing and conversation abilities in this context. ChatEdit is constructed from the CelebA-HQ dataset, incorporating annotated multi-turn dialogues corresponding to user editing requests on the images. The dataset is challenging, as it requires the system to dynamically track and edit images based on user requests, while generating appropriate natural language responses. To address these challenges, we propose a framework comprising a dialogue module for tracking user requests as well as generating responses, and an image editing module for editing images accordingly. Unlike previous approaches, our framework directly tracks the user request of the current turn from the entire dialogue history and edits the initial image instead of manipulating the output from the previous turn, mitigating error accumulation and attribute forgetting issues. Extensive experiments on the ChatEdit dataset demonstrate the superiority of our framework over previous methods and also improvement rooms, encouraging future research. We will release the code and data publicly to facilitate advancements in complex interactive facial image editing. | [
"Cui, Xing",
"Li, Zekun",
"Li, Pei",
"Hu, Yibo",
"Shi, Hailin",
"Cao, Chunshui",
"He, Zhaofeng"
] | ChatEdit: Towards Multi-turn Interactive Facial Image Editing via Dialogue | emnlp-main.899 | 2303.11108 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
|
https://aclanthology.org/2023.emnlp-main.900.bib | https://aclanthology.org/2023.emnlp-main.900/ | @inproceedings{huang-etal-2023-bridging,
title = "Bridging Continuous and Discrete Spaces: Interpretable Sentence Representation Learning via Compositional Operations",
author = "Huang, James Y. and
Yao, Wenlin and
Song, Kaiqiang and
Zhang, Hongming and
Chen, Muhao and
Yu, Dong",
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.900",
doi = "10.18653/v1/2023.emnlp-main.900",
pages = "14584--14595",
abstract = "Traditional sentence embedding models encode sentences into vector representations to capture useful properties such as the semantic similarity between sentences. However, in addition to similarity, sentence semantics can also be interpreted via compositional operations such as sentence fusion or difference. It is unclear whether the compositional semantics of sentences can be directly reflected as compositional operations in the embedding space. To more effectively bridge the continuous embedding and discrete text spaces, we explore the plausibility of incorporating various compositional properties into the sentence embedding space that allows us to interpret embedding transformations as compositional sentence operations. We propose InterSent, an end-to-end framework for learning interpretable sentence embeddings that supports compositional sentence operations in the embedding space. Our method optimizes operator networks and a bottleneck encoder-decoder model to produce meaningful and interpretable sentence embeddings. Experimental results demonstrate that our method significantly improves the interpretability of sentence embeddings on four textual generation tasks over existing approaches while maintaining strong performance on traditional semantic similarity tasks.",
}
| Traditional sentence embedding models encode sentences into vector representations to capture useful properties such as the semantic similarity between sentences. However, in addition to similarity, sentence semantics can also be interpreted via compositional operations such as sentence fusion or difference. It is unclear whether the compositional semantics of sentences can be directly reflected as compositional operations in the embedding space. To more effectively bridge the continuous embedding and discrete text spaces, we explore the plausibility of incorporating various compositional properties into the sentence embedding space that allows us to interpret embedding transformations as compositional sentence operations. We propose InterSent, an end-to-end framework for learning interpretable sentence embeddings that supports compositional sentence operations in the embedding space. Our method optimizes operator networks and a bottleneck encoder-decoder model to produce meaningful and interpretable sentence embeddings. Experimental results demonstrate that our method significantly improves the interpretability of sentence embeddings on four textual generation tasks over existing approaches while maintaining strong performance on traditional semantic similarity tasks. | [
"Huang, James Y.",
"Yao, Wenlin",
"Song, Kaiqiang",
"Zhang, Hongming",
"Chen, Muhao",
"Yu, Dong"
] | Bridging Continuous and Discrete Spaces: Interpretable Sentence Representation Learning via Compositional Operations | emnlp-main.900 | 2305.14599 | [
"https://github.com/jyhuang36/intersent"
] | -1 | -1 | -1 | -1 | [] | [] | [] | 0 | Poster |
Subsets and Splits