bibtex_url
stringlengths 41
50
| proceedings
stringlengths 38
47
| bibtext
stringlengths 709
3.56k
| abstract
stringlengths 17
2.11k
| authors
sequencelengths 1
72
| title
stringlengths 12
207
| id
stringlengths 7
16
| type
stringclasses 2
values | arxiv_id
stringlengths 0
10
| GitHub
sequencelengths 1
1
| paper_page
stringclasses 276
values | n_linked_authors
int64 -1
13
| upvotes
int64 -1
14
| num_comments
int64 -1
11
| n_authors
int64 -1
44
| paper_page_exists_pre_conf
int64 0
1
| Models
sequencelengths 0
100
| Datasets
sequencelengths 0
14
| Spaces
sequencelengths 0
100
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
https://aclanthology.org/2023.acl-long.911.bib | https://aclanthology.org/2023.acl-long.911/ | @inproceedings{rogers-etal-2023-report,
title = "Program Chairs{'} Report on Peer Review at ACL 2023",
author = "Rogers, Anna and
Karpinska, Marzena and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.report",
pages = "xl--lxxv",
abstract = "We present a summary of the efforts to improve conference peer review that were implemented at ACL{'}23. This includes work with the goal of improving review quality, clearer workflow and decision support for the area chairs, as well as our efforts to improve paper-reviewer matching for various kinds of non- mainstream NLP work, and improve the overall incentives for all participants of the peer review process. We present analysis of the factors affecting peer review, identify the most problematic issues that the authors complained about, and provide suggestions for the future chairs. We hope that publishing such reports would (a) improve transparency in decision-making, (b) help the people new to the field to understand how the *ACL conferences work, (c) provide useful data for the future chairs and workshop organizers, and also academic work on peer review, and (d) provide useful context for the final program, as a source of information for meta-research on the structure and trajectory of the field of NLP.",
}
| We present a summary of the efforts to improve conference peer review that were implemented at ACL{'}23. This includes work with the goal of improving review quality, clearer workflow and decision support for the area chairs, as well as our efforts to improve paper-reviewer matching for various kinds of non- mainstream NLP work, and improve the overall incentives for all participants of the peer review process. We present analysis of the factors affecting peer review, identify the most problematic issues that the authors complained about, and provide suggestions for the future chairs. We hope that publishing such reports would (a) improve transparency in decision-making, (b) help the people new to the field to understand how the *ACL conferences work, (c) provide useful data for the future chairs and workshop organizers, and also academic work on peer review, and (d) provide useful context for the final program, as a source of information for meta-research on the structure and trajectory of the field of NLP. | [
"Rogers, Anna",
"Karpinska, Marzena",
"Boyd-Graber, Jordan",
"Okazaki, Naoaki"
] | Program Chairs' Report on Peer Review at ACL 2023 | acl-long.911 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.1.bib | https://aclanthology.org/2023.acl-long.1/ | @inproceedings{liu-etal-2023-one,
title = "One Cannot Stand for Everyone! Leveraging Multiple User Simulators to train Task-oriented Dialogue Systems",
author = "Liu, Yajiao and
Jiang, Xin and
Yin, Yichun and
Wang, Yasheng and
Mi, Fei and
Liu, Qun and
Wan, Xiang and
Wang, Benyou",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.1",
doi = "10.18653/v1/2023.acl-long.1",
pages = "1--21",
abstract = "User simulators are agents designed to imitate human users; recent advances have found that Task-oriented Dialogue (ToD) systems optimized toward a user simulator could better satisfy the need of human users. However, this might result in a sub-optimal ToD system if it is tailored to only one \textit{ad hoc} user simulator, since human users can behave differently. In this paper, we propose a framework called MUST to optimize ToD systems via leveraging Multiple User SimulaTors. The main challenges of implementing MUST fall in 1) how to adaptively determine which user simulator to interact with the ToD system at each optimization step, since the ToD system might be over-fitted to some specific user simulators, and simultaneously under-fitted to some others; 2) how to avoid catastrophic forgetting of the adaption for a simulator that is not selected for several consecutive optimization steps.To tackle these challenges, we formulate MUST as a Multi-armed bandits (MAB) problem and provide a method called MUST$_{\mathrm{adaptive}}$ that balances \textit{i}) the \textit{boosting adaption} for adaptive interactions between different user simulators and the ToD system and\textit{ii}) the \textit{uniform adaption} to avoid the catastrophic forgetting issue.With both automatic evaluations and human evaluations, our experimental results on MultiWOZ show that the dialogue system trained by MUST achieves a better performance than those trained by a single user simulator. It also has a better generalization ability when testing with unseen user simulators.",
}
| User simulators are agents designed to imitate human users; recent advances have found that Task-oriented Dialogue (ToD) systems optimized toward a user simulator could better satisfy the need of human users. However, this might result in a sub-optimal ToD system if it is tailored to only one \textit{ad hoc} user simulator, since human users can behave differently. In this paper, we propose a framework called MUST to optimize ToD systems via leveraging Multiple User SimulaTors. The main challenges of implementing MUST fall in 1) how to adaptively determine which user simulator to interact with the ToD system at each optimization step, since the ToD system might be over-fitted to some specific user simulators, and simultaneously under-fitted to some others; 2) how to avoid catastrophic forgetting of the adaption for a simulator that is not selected for several consecutive optimization steps.To tackle these challenges, we formulate MUST as a Multi-armed bandits (MAB) problem and provide a method called MUST$_{\mathrm{adaptive}}$ that balances \textit{i}) the \textit{boosting adaption} for adaptive interactions between different user simulators and the ToD system and\textit{ii}) the \textit{uniform adaption} to avoid the catastrophic forgetting issue.With both automatic evaluations and human evaluations, our experimental results on MultiWOZ show that the dialogue system trained by MUST achieves a better performance than those trained by a single user simulator. It also has a better generalization ability when testing with unseen user simulators. | [
"Liu, Yajiao",
"Jiang, Xin",
"Yin, Yichun",
"Wang, Yasheng",
"Mi, Fei",
"Liu, Qun",
"Wan, Xiang",
"Wang, Benyou"
] | One Cannot Stand for Everyone! Leveraging Multiple User Simulators to train Task-oriented Dialogue Systems | acl-long.1 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.2.bib | https://aclanthology.org/2023.acl-long.2/ | @inproceedings{zhang-etal-2023-safeconv,
title = "{S}afe{C}onv: Explaining and Correcting Conversational Unsafe Behavior",
author = "Zhang, Mian and
Jin, Lifeng and
Song, Linfeng and
Mi, Haitao and
Chen, Wenliang and
Yu, Dong",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.2",
doi = "10.18653/v1/2023.acl-long.2",
pages = "22--35",
abstract = "One of the main challenges open-domain end-to-end dialogue systems, or chatbots, face is the prevalence of unsafe behavior, such as toxic languages and harmful suggestions. However, existing dialogue datasets do not provide enough annotation to explain and correct such unsafe behavior. In this work, we construct a new dataset called SafeConv for the research of conversational safety: (1) Besides the utterance-level safety labels, SafeConv also provides unsafe spans in an utterance, information able to indicate which words contribute to the detected unsafe behavior; (2) SafeConv provides safe alternative responses to continue the conversation when unsafe behavior detected, guiding the conversation to a gentle trajectory. By virtue of the comprehensive annotation of SafeConv, we benchmark three powerful models for the mitigation of conversational unsafe behavior, including a checker to detect unsafe utterances, a tagger to extract unsafe spans, and a rewriter to convert an unsafe response to a safe version. Moreover, we explore the huge benefits brought by combining the models for explaining the emergence of unsafe behavior and detoxifying chatbots. Experiments show that the detected unsafe behavior could be well explained with unsafe spans and popular chatbots could be detoxified by a huge extent. The dataset is available at \url{https://github.com/mianzhang/SafeConv}.",
}
| One of the main challenges open-domain end-to-end dialogue systems, or chatbots, face is the prevalence of unsafe behavior, such as toxic languages and harmful suggestions. However, existing dialogue datasets do not provide enough annotation to explain and correct such unsafe behavior. In this work, we construct a new dataset called SafeConv for the research of conversational safety: (1) Besides the utterance-level safety labels, SafeConv also provides unsafe spans in an utterance, information able to indicate which words contribute to the detected unsafe behavior; (2) SafeConv provides safe alternative responses to continue the conversation when unsafe behavior detected, guiding the conversation to a gentle trajectory. By virtue of the comprehensive annotation of SafeConv, we benchmark three powerful models for the mitigation of conversational unsafe behavior, including a checker to detect unsafe utterances, a tagger to extract unsafe spans, and a rewriter to convert an unsafe response to a safe version. Moreover, we explore the huge benefits brought by combining the models for explaining the emergence of unsafe behavior and detoxifying chatbots. Experiments show that the detected unsafe behavior could be well explained with unsafe spans and popular chatbots could be detoxified by a huge extent. The dataset is available at \url{https://github.com/mianzhang/SafeConv}. | [
"Zhang, Mian",
"Jin, Lifeng",
"Song, Linfeng",
"Mi, Haitao",
"Chen, Wenliang",
"Yu, Dong"
] | SafeConv: Explaining and Correcting Conversational Unsafe Behavior | acl-long.2 | Oral | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.3.bib | https://aclanthology.org/2023.acl-long.3/ | @inproceedings{dale-etal-2023-detecting,
title = "Detecting and Mitigating Hallucinations in Machine Translation: Model Internal Workings Alone Do Well, Sentence Similarity {E}ven Better",
author = "Dale, David and
Voita, Elena and
Barrault, Loic and
Costa-juss{\`a}, Marta R.",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.3",
doi = "10.18653/v1/2023.acl-long.3",
pages = "36--50",
abstract = "While the problem of hallucinations in neural machine translation has long been recognized, so far the progress on its alleviation is very little. Indeed, recently it turned out that without artificially encouraging models to hallucinate, previously existing methods fall short and even the standard sequence log-probability is more informative. It means that internal characteristics of the model can give much more information than we expect, and before using external models and measures, we first need to ask: how far can we go if we use nothing but the translation model itself ? We propose to use a method that evaluates the percentage of the source contribution to a generated translation. Intuitively, hallucinations are translations {``}detached{''} from the source, hence they can be identified by low source contribution. This method improves detection accuracy for the most severe hallucinations by a factor of 2 and is able to alleviate hallucinations at test time on par with the previous best approach that relies on external models. Next, if we move away from internal model characteristics and allow external tools, we show that using sentence similarity from cross-lingual embeddings further improves these results. We release the code of our experiments.",
}
| While the problem of hallucinations in neural machine translation has long been recognized, so far the progress on its alleviation is very little. Indeed, recently it turned out that without artificially encouraging models to hallucinate, previously existing methods fall short and even the standard sequence log-probability is more informative. It means that internal characteristics of the model can give much more information than we expect, and before using external models and measures, we first need to ask: how far can we go if we use nothing but the translation model itself ? We propose to use a method that evaluates the percentage of the source contribution to a generated translation. Intuitively, hallucinations are translations {``}detached{''} from the source, hence they can be identified by low source contribution. This method improves detection accuracy for the most severe hallucinations by a factor of 2 and is able to alleviate hallucinations at test time on par with the previous best approach that relies on external models. Next, if we move away from internal model characteristics and allow external tools, we show that using sentence similarity from cross-lingual embeddings further improves these results. We release the code of our experiments. | [
"Dale, David",
"Voita, Elena",
"Barrault, Loic",
"Costa-juss{\\`a}, Marta R."
] | Detecting and Mitigating Hallucinations in Machine Translation: Model Internal Workings Alone Do Well, Sentence Similarity Even Better | acl-long.3 | Poster | 2212.08597 | [
""
] | https://huggingface.co./papers/2212.08597 | 1 | 1 | 0 | 4 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.4.bib | https://aclanthology.org/2023.acl-long.4/ | @inproceedings{cheng-etal-2023-explainable,
title = "Explainable Recommendation with Personalized Review Retrieval and Aspect Learning",
author = "Cheng, Hao and
Wang, Shuo and
Lu, Wensheng and
Zhang, Wei and
Zhou, Mingyang and
Lu, Kezhong and
Liao, Hao",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.4",
doi = "10.18653/v1/2023.acl-long.4",
pages = "51--64",
abstract = "Explainable recommendation is a technique that combines prediction and generation tasks to produce more persuasive results. Among these tasks, textual generation demands large amounts of data to achieve satisfactory accuracy. However, historical user reviews of items are often insufficient, making it challenging to ensure the precision of generated explanation text. To address this issue, we propose a novel model, ERRA (Explainable Recommendation by personalized Review retrieval and Aspect learning). With retrieval enhancement, ERRA can obtain additional information from the training sets. With this additional information, we can generate more accurate and informative explanations. Furthermore, to better capture users{'} preferences, we incorporate an aspect enhancement component into our model. By selecting the top-n aspects that users are most concerned about for different items, we can model user representation with more relevant details, making the explanation more persuasive. To verify the effectiveness of our model, extensive experiments on three datasets show that our model outperforms state-of-the-art baselines (for example, 3.4{\%} improvement in prediction and 15.8{\%} improvement in explanation for TripAdvisor).",
}
| Explainable recommendation is a technique that combines prediction and generation tasks to produce more persuasive results. Among these tasks, textual generation demands large amounts of data to achieve satisfactory accuracy. However, historical user reviews of items are often insufficient, making it challenging to ensure the precision of generated explanation text. To address this issue, we propose a novel model, ERRA (Explainable Recommendation by personalized Review retrieval and Aspect learning). With retrieval enhancement, ERRA can obtain additional information from the training sets. With this additional information, we can generate more accurate and informative explanations. Furthermore, to better capture users{'} preferences, we incorporate an aspect enhancement component into our model. By selecting the top-n aspects that users are most concerned about for different items, we can model user representation with more relevant details, making the explanation more persuasive. To verify the effectiveness of our model, extensive experiments on three datasets show that our model outperforms state-of-the-art baselines (for example, 3.4{\%} improvement in prediction and 15.8{\%} improvement in explanation for TripAdvisor). | [
"Cheng, Hao",
"Wang, Shuo",
"Lu, Wensheng",
"Zhang, Wei",
"Zhou, Mingyang",
"Lu, Kezhong",
"Liao, Hao"
] | Explainable Recommendation with Personalized Review Retrieval and Aspect Learning | acl-long.4 | Poster | 2306.12657 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.5.bib | https://aclanthology.org/2023.acl-long.5/ | @inproceedings{liu-etal-2023-binary,
title = "Binary and Ternary Natural Language Generation",
author = "Liu, Zechun and
Oguz, Barlas and
Pappu, Aasish and
Shi, Yangyang and
Krishnamoorthi, Raghuraman",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.5",
doi = "10.18653/v1/2023.acl-long.5",
pages = "65--77",
abstract = "Ternary and binary neural networks enable multiplication-free computation and promise multiple orders of magnitude efficiency gains over full-precision networks if implemented on specialized hardware. However, since both the parameter and the output space are highly discretized, such networks have proven very difficult to optimize. The difficulties are compounded for the class of transformer text generation models due to the sensitivity of the attention operation to quantization and the noise-compounding effects of autoregressive decoding in the high-cardinality output space. We approach the problem with a mix of statistics-based quantization for the weights and elastic quantization of the activations and demonstrate the first ternary and binary transformer models on the downstream tasks of summarization and machine translation. Our ternary BART base achieves an R1 score of 41 on the CNN/DailyMail benchmark, which is merely 3.9 points behind the full model while being 16x more efficient. Our binary model, while less accurate, achieves a highly non-trivial score of 35.6. For machine translation, we achieved BLEU scores of 21.7 and 17.6 on the WMT16 En-Ro benchmark, compared with a full precision mBART model score of 26.8. We also compare our approach in the 8-bit activation setting, where our ternary and even binary weight models can match or outperform the best existing 8-bit weight models in the literature. Our code and models are available at: \url{https://github.com/facebookresearch/Ternary_Binary_Transformer}.",
}
| Ternary and binary neural networks enable multiplication-free computation and promise multiple orders of magnitude efficiency gains over full-precision networks if implemented on specialized hardware. However, since both the parameter and the output space are highly discretized, such networks have proven very difficult to optimize. The difficulties are compounded for the class of transformer text generation models due to the sensitivity of the attention operation to quantization and the noise-compounding effects of autoregressive decoding in the high-cardinality output space. We approach the problem with a mix of statistics-based quantization for the weights and elastic quantization of the activations and demonstrate the first ternary and binary transformer models on the downstream tasks of summarization and machine translation. Our ternary BART base achieves an R1 score of 41 on the CNN/DailyMail benchmark, which is merely 3.9 points behind the full model while being 16x more efficient. Our binary model, while less accurate, achieves a highly non-trivial score of 35.6. For machine translation, we achieved BLEU scores of 21.7 and 17.6 on the WMT16 En-Ro benchmark, compared with a full precision mBART model score of 26.8. We also compare our approach in the 8-bit activation setting, where our ternary and even binary weight models can match or outperform the best existing 8-bit weight models in the literature. Our code and models are available at: \url{https://github.com/facebookresearch/Ternary_Binary_Transformer}. | [
"Liu, Zechun",
"Oguz, Barlas",
"Pappu, Aasish",
"Shi, Yangyang",
"Krishnamoorthi, Raghuraman"
] | Binary and Ternary Natural Language Generation | acl-long.5 | Oral | 2306.01841 | [
"https://github.com/facebookresearch/ternary_binary_transformer"
] | https://huggingface.co./papers/2306.01841 | 1 | 2 | 0 | 5 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.6.bib | https://aclanthology.org/2023.acl-long.6/ | @inproceedings{bebensee-lee-2023-span,
title = "Span-Selective Linear Attention Transformers for Effective and Robust Schema-Guided Dialogue State Tracking",
author = {Bebensee, Bj{\"o}rn and
Lee, Haejun},
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.6",
doi = "10.18653/v1/2023.acl-long.6",
pages = "78--91",
abstract = "In schema-guided dialogue state tracking models estimate the current state of a conversation using natural language descriptions of the service schema for generalization to unseen services. Prior generative approaches which decode slot values sequentially do not generalize well to variations in schema, while discriminative approaches separately encode history and schema and fail to account for inter-slot and intent-slot dependencies. We introduce SPLAT, a novel architecture which achieves better generalization and efficiency than prior approaches by constraining outputs to a limited prediction space. At the same time, our model allows for rich attention among descriptions and history while keeping computation costs constrained by incorporating linear-time attention. We demonstrate the effectiveness of our model on the Schema-Guided Dialogue (SGD) and MultiWOZ datasets. Our approach significantly improves upon existing models achieving 85.3 JGA on the SGD dataset. Further, we show increased robustness on the SGD-X benchmark: our model outperforms the more than 30x larger D3ST-XXL model by 5.0 points.",
}
| In schema-guided dialogue state tracking models estimate the current state of a conversation using natural language descriptions of the service schema for generalization to unseen services. Prior generative approaches which decode slot values sequentially do not generalize well to variations in schema, while discriminative approaches separately encode history and schema and fail to account for inter-slot and intent-slot dependencies. We introduce SPLAT, a novel architecture which achieves better generalization and efficiency than prior approaches by constraining outputs to a limited prediction space. At the same time, our model allows for rich attention among descriptions and history while keeping computation costs constrained by incorporating linear-time attention. We demonstrate the effectiveness of our model on the Schema-Guided Dialogue (SGD) and MultiWOZ datasets. Our approach significantly improves upon existing models achieving 85.3 JGA on the SGD dataset. Further, we show increased robustness on the SGD-X benchmark: our model outperforms the more than 30x larger D3ST-XXL model by 5.0 points. | [
"Bebensee, Bj{\\\"o}rn",
"Lee, Haejun"
] | Span-Selective Linear Attention Transformers for Effective and Robust Schema-Guided Dialogue State Tracking | acl-long.6 | Poster | 2306.09340 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.7.bib | https://aclanthology.org/2023.acl-long.7/ | @inproceedings{li-zhao-2023-em,
title = "{EM} Pre-training for Multi-party Dialogue Response Generation",
author = "Li, Yiyang and
Zhao, Hai",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.7",
doi = "10.18653/v1/2023.acl-long.7",
pages = "92--103",
abstract = "Dialogue response generation requires an agent to generate a response according to the current dialogue history, in terms of which two-party dialogues have been well studied, but leaving a great gap for multi-party dialogues at the same time. Different from two-party dialogues where each response is a direct reply to its previous utterance, the addressee of a response utterance should be specified before it is generated in the multi-party scenario. Thanks to the huge amount of two-party conversational data, various pre-trained language models for two-party dialogue response generation have been proposed. However, due to the lack of annotated addressee labels in multi-party dialogue datasets, it is hard to use them to pre-train a response generation model for multi-party dialogues. To tackle this obstacle, we propose an Expectation-Maximization (EM) approach that iteratively performs the expectation steps to generate addressee labels, and the maximization steps to optimize a response generation model. Theoretical analyses and extensive experiments have justified the feasibility and effectiveness of our proposed method. The official implementation of this paper is available at \url{https://github.com/EricLee8/MPDRG}.",
}
| Dialogue response generation requires an agent to generate a response according to the current dialogue history, in terms of which two-party dialogues have been well studied, but leaving a great gap for multi-party dialogues at the same time. Different from two-party dialogues where each response is a direct reply to its previous utterance, the addressee of a response utterance should be specified before it is generated in the multi-party scenario. Thanks to the huge amount of two-party conversational data, various pre-trained language models for two-party dialogue response generation have been proposed. However, due to the lack of annotated addressee labels in multi-party dialogue datasets, it is hard to use them to pre-train a response generation model for multi-party dialogues. To tackle this obstacle, we propose an Expectation-Maximization (EM) approach that iteratively performs the expectation steps to generate addressee labels, and the maximization steps to optimize a response generation model. Theoretical analyses and extensive experiments have justified the feasibility and effectiveness of our proposed method. The official implementation of this paper is available at \url{https://github.com/EricLee8/MPDRG}. | [
"Li, Yiyang",
"Zhao, Hai"
] | EM Pre-training for Multi-party Dialogue Response Generation | acl-long.7 | Poster | 2305.12412 | [
"https://github.com/ericlee8/mpdrg"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.8.bib | https://aclanthology.org/2023.acl-long.8/ | @inproceedings{ghosh-etal-2023-aclm,
title = "{ACLM}: A Selective-Denoising based Generative Data Augmentation Approach for Low-Resource Complex {NER}",
author = "Ghosh, Sreyan and
Tyagi, Utkarsh and
Suri, Manan and
Kumar, Sonal and
S, Ramaneswaran and
Manocha, Dinesh",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.8",
doi = "10.18653/v1/2023.acl-long.8",
pages = "104--125",
abstract = "Complex Named Entity Recognition (NER) is the task of detecting linguistically complex named entities in low-context text. In this paper, we present ACLM Attention-map aware keyword selection for Conditional Language Model fine-tuning), a novel data augmentation approach based on conditional generation, to address the data scarcity problem in low-resource complex NER. ACLM alleviates the context-entity mismatch issue, a problem existing NER data augmentation techniques suffer from and often generates incoherent augmentations by placing complex named entities in the wrong context. ACLM builds on BART and is optimized on a novel text reconstruction or denoising task - we use selective masking (aided by attention maps) to retain the named entities and certain keywords in the input sentence that provide contextually relevant additional knowledge or hints about the named entities. Compared with other data augmentation strategies, ACLM can generate more diverse and coherent augmentations preserving the true word sense of complex entities in the sentence. We demonstrate the effectiveness of ACLM both qualitatively and quantitatively on monolingual, cross-lingual, and multilingual complex NER across various low-resource settings. ACLM outperforms all our neural baselines by a significant margin (1{\%}-36{\%}). In addition, we demonstrate the application of ACLM to other domains that suffer from data scarcity (e.g., biomedical). In practice, ACLM generates more effective and factual augmentations for these domains than prior methods.",
}
| Complex Named Entity Recognition (NER) is the task of detecting linguistically complex named entities in low-context text. In this paper, we present ACLM Attention-map aware keyword selection for Conditional Language Model fine-tuning), a novel data augmentation approach based on conditional generation, to address the data scarcity problem in low-resource complex NER. ACLM alleviates the context-entity mismatch issue, a problem existing NER data augmentation techniques suffer from and often generates incoherent augmentations by placing complex named entities in the wrong context. ACLM builds on BART and is optimized on a novel text reconstruction or denoising task - we use selective masking (aided by attention maps) to retain the named entities and certain keywords in the input sentence that provide contextually relevant additional knowledge or hints about the named entities. Compared with other data augmentation strategies, ACLM can generate more diverse and coherent augmentations preserving the true word sense of complex entities in the sentence. We demonstrate the effectiveness of ACLM both qualitatively and quantitatively on monolingual, cross-lingual, and multilingual complex NER across various low-resource settings. ACLM outperforms all our neural baselines by a significant margin (1{\%}-36{\%}). In addition, we demonstrate the application of ACLM to other domains that suffer from data scarcity (e.g., biomedical). In practice, ACLM generates more effective and factual augmentations for these domains than prior methods. | [
"Ghosh, Sreyan",
"Tyagi, Utkarsh",
"Suri, Manan",
"Kumar, Sonal",
"S, Ramaneswaran",
"Manocha, Dinesh"
] | ACLM: A Selective-Denoising based Generative Data Augmentation Approach for Low-Resource Complex NER | acl-long.8 | Poster | 2306.00928 | [
"https://github.com/sreyan88/aclm"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.9.bib | https://aclanthology.org/2023.acl-long.9/ | @inproceedings{yin-etal-2023-natural,
title = "Natural Language to Code Generation in Interactive Data Science Notebooks",
author = "Yin, Pengcheng and
Li, Wen-Ding and
Xiao, Kefan and
Rao, Abhishek and
Wen, Yeming and
Shi, Kensen and
Howland, Joshua and
Bailey, Paige and
Catasta, Michele and
Michalewski, Henryk and
Polozov, Oleksandr and
Sutton, Charles",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.9",
doi = "10.18653/v1/2023.acl-long.9",
pages = "126--173",
abstract = "Computational notebooks, such as Jupyter notebooks, are interactive computing environments that are ubiquitous among data scientists to perform data wrangling and analytic tasks. To measure the performance of AI pair programmers that automatically synthesize programs for those tasks given natural language (NL) intents from users, we build ARCADE, a benchmark of 1078 code generation problems using the pandas data analysis framework in data science notebooks. ARCADE features multiple rounds of NL-to-code problems from the same notebook. It requires a model to understand rich multi-modal contexts, such as existing notebook cells and their execution states as well as previous turns of interaction. To establish a strong baseline on this challenging task, we develop PaChiNCo, a 62B code language model (LM) for Python computational notebooks, which significantly outperforms public code LMs. Finally, we explore few-shot prompting strategies to elicit better code with step-by-step decomposition and NL explanation, showing the potential to improve the diversity and explainability of model predictions. Arcade is publicly available at \url{https://github.com/google-research/arcade-nl2code/}.",
}
| Computational notebooks, such as Jupyter notebooks, are interactive computing environments that are ubiquitous among data scientists to perform data wrangling and analytic tasks. To measure the performance of AI pair programmers that automatically synthesize programs for those tasks given natural language (NL) intents from users, we build ARCADE, a benchmark of 1078 code generation problems using the pandas data analysis framework in data science notebooks. ARCADE features multiple rounds of NL-to-code problems from the same notebook. It requires a model to understand rich multi-modal contexts, such as existing notebook cells and their execution states as well as previous turns of interaction. To establish a strong baseline on this challenging task, we develop PaChiNCo, a 62B code language model (LM) for Python computational notebooks, which significantly outperforms public code LMs. Finally, we explore few-shot prompting strategies to elicit better code with step-by-step decomposition and NL explanation, showing the potential to improve the diversity and explainability of model predictions. Arcade is publicly available at \url{https://github.com/google-research/arcade-nl2code/}. | [
"Yin, Pengcheng",
"Li, Wen-Ding",
"Xiao, Kefan",
"Rao, Abhishek",
"Wen, Yeming",
"Shi, Kensen",
"Howl",
", Joshua",
"Bailey, Paige",
"Catasta, Michele",
"Michalewski, Henryk",
"Polozov, Oleks",
"r",
"Sutton, Charles"
] | Natural Language to Code Generation in Interactive Data Science Notebooks | acl-long.9 | Poster | 2212.09248 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.10.bib | https://aclanthology.org/2023.acl-long.10/ | @inproceedings{deguchi-etal-2023-subset,
title = "Subset Retrieval Nearest Neighbor Machine Translation",
author = "Deguchi, Hiroyuki and
Watanabe, Taro and
Matsui, Yusuke and
Utiyama, Masao and
Tanaka, Hideki and
Sumita, Eiichiro",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.10",
doi = "10.18653/v1/2023.acl-long.10",
pages = "174--189",
abstract = "k-nearest-neighbor machine translation (kNN-MT) (Khandelwal et al., 2021) boosts the translation performance of trained neural machine translation (NMT) models by incorporating example-search into the decoding algorithm. However, decoding is seriously time-consuming, i.e., roughly 100 to 1,000 times slower than standard NMT, because neighbor tokens are retrieved from all target tokens of parallel data in each timestep. In this paper, we propose {``}Subset kNN-MT{''}, which improves the decoding speed of kNN-MT by two methods: (1) retrieving neighbor target tokens from a subset that is the set of neighbor sentences of the input sentence, not from all sentences, and (2) efficient distance computation technique that is suitable for subset neighbor search using a look-up table. Our proposed method achieved a speed-up of up to 132.2 times and an improvement in BLEU score of up to 1.6 compared with kNN-MT in the WMT{'}19 De-En translation task and the domain adaptation tasks in De-En and En-Ja.",
}
| k-nearest-neighbor machine translation (kNN-MT) (Khandelwal et al., 2021) boosts the translation performance of trained neural machine translation (NMT) models by incorporating example-search into the decoding algorithm. However, decoding is seriously time-consuming, i.e., roughly 100 to 1,000 times slower than standard NMT, because neighbor tokens are retrieved from all target tokens of parallel data in each timestep. In this paper, we propose {``}Subset kNN-MT{''}, which improves the decoding speed of kNN-MT by two methods: (1) retrieving neighbor target tokens from a subset that is the set of neighbor sentences of the input sentence, not from all sentences, and (2) efficient distance computation technique that is suitable for subset neighbor search using a look-up table. Our proposed method achieved a speed-up of up to 132.2 times and an improvement in BLEU score of up to 1.6 compared with kNN-MT in the WMT{'}19 De-En translation task and the domain adaptation tasks in De-En and En-Ja. | [
"Deguchi, Hiroyuki",
"Watanabe, Taro",
"Matsui, Yusuke",
"Utiyama, Masao",
"Tanaka, Hideki",
"Sumita, Eiichiro"
] | Subset Retrieval Nearest Neighbor Machine Translation | acl-long.10 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.11.bib | https://aclanthology.org/2023.acl-long.11/ | @inproceedings{zhang-wan-2023-mil,
title = "{MIL}-Decoding: Detoxifying Language Models at Token-Level via Multiple Instance Learning",
author = "Zhang, Xu and
Wan, Xiaojun",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.11",
doi = "10.18653/v1/2023.acl-long.11",
pages = "190--202",
abstract = "Despite advances in large pre-trained neural language models, they are prone to generating toxic language, which brings security risks to their applications. We introduce MIL-Decoding, which detoxifies language models at token-level by interpolating it with a trained multiple instance learning (MIL) network.MIL model is trained on a corpus with a toxicity label for each text to predict the overall toxicity and the toxicity of each token in its context. Intuitively, the MIL network computes a toxicity distribution over next tokens according to the generated context which supplements the original language model to avoid toxicity. We evaluate MIL-Decoding with automatic metrics and human evaluation, where MIL-Decoding outperforms other baselines in detoxification while it only hurts generation fluency a little bit.",
}
| Despite advances in large pre-trained neural language models, they are prone to generating toxic language, which brings security risks to their applications. We introduce MIL-Decoding, which detoxifies language models at token-level by interpolating it with a trained multiple instance learning (MIL) network.MIL model is trained on a corpus with a toxicity label for each text to predict the overall toxicity and the toxicity of each token in its context. Intuitively, the MIL network computes a toxicity distribution over next tokens according to the generated context which supplements the original language model to avoid toxicity. We evaluate MIL-Decoding with automatic metrics and human evaluation, where MIL-Decoding outperforms other baselines in detoxification while it only hurts generation fluency a little bit. | [
"Zhang, Xu",
"Wan, Xiaojun"
] | MIL-Decoding: Detoxifying Language Models at Token-Level via Multiple Instance Learning | acl-long.11 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.12.bib | https://aclanthology.org/2023.acl-long.12/ | @inproceedings{de-dios-flores-etal-2023-dependency,
title = "Dependency resolution at the syntax-semantics interface: psycholinguistic and computational insights on control dependencies",
author = "de-Dios-Flores, Iria and
Garcia Amboage, Juan and
Garcia, Marcos",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.12",
doi = "10.18653/v1/2023.acl-long.12",
pages = "203--222",
abstract = "Using psycholinguistic and computational experiments we compare the ability of humans and several pre-trained masked language models to correctly identify control dependencies in Spanish sentences such as {`}Jos{\'e} le prometi{\'o}/orden{\'o} a Mar{\'\i}a ser ordenado/a{'} ({`}Joseph promised/ordered Mary to be tidy{'}). These structures underlie complex anaphoric and agreement relations at the interface of syntax and semantics, allowing us to study lexically-guided antecedent retrieval processes. Our results show that while humans correctly identify the (un)acceptability of the strings, language models often fail to identify the correct antecedent in non-adjacent dependencies, showing their reliance on linearity. Additional experiments on Galician reinforce these conclusions. Our findings are equally valuable for the evaluation of language models{'} ability to capture linguistic generalizations, as well as for psycholinguistic theories of anaphor resolution.",
}
| Using psycholinguistic and computational experiments we compare the ability of humans and several pre-trained masked language models to correctly identify control dependencies in Spanish sentences such as {`}Jos{\'e} le prometi{\'o}/orden{\'o} a Mar{\'\i}a ser ordenado/a{'} ({`}Joseph promised/ordered Mary to be tidy{'}). These structures underlie complex anaphoric and agreement relations at the interface of syntax and semantics, allowing us to study lexically-guided antecedent retrieval processes. Our results show that while humans correctly identify the (un)acceptability of the strings, language models often fail to identify the correct antecedent in non-adjacent dependencies, showing their reliance on linearity. Additional experiments on Galician reinforce these conclusions. Our findings are equally valuable for the evaluation of language models{'} ability to capture linguistic generalizations, as well as for psycholinguistic theories of anaphor resolution. | [
"de-Dios-Flores, Iria",
"Garcia Amboage, Juan",
"Garcia, Marcos"
] | Dependency resolution at the syntax-semantics interface: psycholinguistic and computational insights on control dependencies | acl-long.12 | Oral | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.13.bib | https://aclanthology.org/2023.acl-long.13/ | @inproceedings{liang-etal-2023-open,
title = "Open-ended Long Text Generation via Masked Language Modeling",
author = "Liang, Xiaobo and
Tang, Zecheng and
Li, Juntao and
Zhang, Min",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.13",
doi = "10.18653/v1/2023.acl-long.13",
pages = "223--241",
abstract = "Pre-trained autoregressive (AR) language models such as BART and GPTs have dominated OPen-ended Long Text Generation (Open-LTG).However, the AR nature will decrease the inference efficiency along with the increase of generation length, which hinder their application in Open-LTG.To improve inference efficiency, we alternatively explore the potential of the pre-trained masked language models (MLMs) along with a representative iterative non-autoregressive (NAR) decoding strategy for Open-LTG.Our preliminary study shows that pre-trained MLMs can merely generate short text and will collapse for long text modeling. To enhance the long text generation capability of MLMs, we introduce two simple yet effective strategies for the iterative NAR model: dynamic sliding window attention (DSWA) and linear temperature decay (LTD). It can alleviate long-distance collapse problems and achieve longer text generation with a flexible trade-off between performance and inference speedup. Experiments on the storytelling and multi-paragraph opinionated article writing tasks show that pre-trained MLMs can achieve more than 3 $\times$ $\to$ 13 $\times$ speedup with better performance than strong AR models.",
}
| Pre-trained autoregressive (AR) language models such as BART and GPTs have dominated OPen-ended Long Text Generation (Open-LTG).However, the AR nature will decrease the inference efficiency along with the increase of generation length, which hinder their application in Open-LTG.To improve inference efficiency, we alternatively explore the potential of the pre-trained masked language models (MLMs) along with a representative iterative non-autoregressive (NAR) decoding strategy for Open-LTG.Our preliminary study shows that pre-trained MLMs can merely generate short text and will collapse for long text modeling. To enhance the long text generation capability of MLMs, we introduce two simple yet effective strategies for the iterative NAR model: dynamic sliding window attention (DSWA) and linear temperature decay (LTD). It can alleviate long-distance collapse problems and achieve longer text generation with a flexible trade-off between performance and inference speedup. Experiments on the storytelling and multi-paragraph opinionated article writing tasks show that pre-trained MLMs can achieve more than 3 $\times$ $\to$ 13 $\times$ speedup with better performance than strong AR models. | [
"Liang, Xiaobo",
"Tang, Zecheng",
"Li, Juntao",
"Zhang, Min"
] | Open-ended Long Text Generation via Masked Language Modeling | acl-long.13 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.14.bib | https://aclanthology.org/2023.acl-long.14/ | @inproceedings{chronis-etal-2023-method,
title = "A Method for Studying Semantic Construal in Grammatical Constructions with Interpretable Contextual Embedding Spaces",
author = "Chronis, Gabriella and
Mahowald, Kyle and
Erk, Katrin",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.14",
doi = "10.18653/v1/2023.acl-long.14",
pages = "242--261",
abstract = "We study semantic construal in grammatical constructions using large language models. First, we project contextual word embeddings into three interpretable semantic spaces, each defined by a different set of psycholinguistic feature norms. We validate these interpretable spaces and then use them to automatically derive semantic characterizations of lexical items in two grammatical constructions: nouns in subject or object position within the same sentence, and the AANN construction (e.g., {`}a beautiful three days{'}). We show that a word in subject position is interpreted as more agentive than the very same word in object position, and that the nouns in the AANN construction are interpreted as more measurement-like than when in the canonical alternation. Our method can probe the distributional meaning of syntactic constructions at a templatic level, abstracted away from specific lexemes.",
}
| We study semantic construal in grammatical constructions using large language models. First, we project contextual word embeddings into three interpretable semantic spaces, each defined by a different set of psycholinguistic feature norms. We validate these interpretable spaces and then use them to automatically derive semantic characterizations of lexical items in two grammatical constructions: nouns in subject or object position within the same sentence, and the AANN construction (e.g., {`}a beautiful three days{'}). We show that a word in subject position is interpreted as more agentive than the very same word in object position, and that the nouns in the AANN construction are interpreted as more measurement-like than when in the canonical alternation. Our method can probe the distributional meaning of syntactic constructions at a templatic level, abstracted away from specific lexemes. | [
"Chronis, Gabriella",
"Mahowald, Kyle",
"Erk, Katrin"
] | A Method for Studying Semantic Construal in Grammatical Constructions with Interpretable Contextual Embedding Spaces | acl-long.14 | Oral | 2305.18598 | [
"https://github.com/gchronis/features_in_context"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.15.bib | https://aclanthology.org/2023.acl-long.15/ | @inproceedings{yamaki-etal-2023-holographic,
title = "Holographic {CCG} Parsing",
author = "Yamaki, Ryosuke and
Taniguchi, Tadahiro and
Mochihashi, Daichi",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.15",
doi = "10.18653/v1/2023.acl-long.15",
pages = "262--276",
abstract = "We propose a method for formulating CCG as a recursive composition in a continuous vector space. Recent CCG supertagging and parsing models generally demonstrate high performance, yet rely on black-box neural architectures to implicitly model phrase structure dependencies. Instead, we leverage the method of holographic embeddings as a compositional operator to explicitly model the dependencies between words and phrase structures in the embedding space. Experimental results revealed that holographic composition effectively improves the supertagging accuracy to achieve state-of-the-art parsing performance when using a C{\&}C parser. The proposed span-based parsing algorithm using holographic composition achieves performance comparable to state-of-the-art neural parsing with Transformers. Furthermore, our model can semantically and syntactically infill text at the phrase level due to the decomposability of holographic composition.",
}
| We propose a method for formulating CCG as a recursive composition in a continuous vector space. Recent CCG supertagging and parsing models generally demonstrate high performance, yet rely on black-box neural architectures to implicitly model phrase structure dependencies. Instead, we leverage the method of holographic embeddings as a compositional operator to explicitly model the dependencies between words and phrase structures in the embedding space. Experimental results revealed that holographic composition effectively improves the supertagging accuracy to achieve state-of-the-art parsing performance when using a C{\&}C parser. The proposed span-based parsing algorithm using holographic composition achieves performance comparable to state-of-the-art neural parsing with Transformers. Furthermore, our model can semantically and syntactically infill text at the phrase level due to the decomposability of holographic composition. | [
"Yamaki, Ryosuke",
"Taniguchi, Tadahiro",
"Mochihashi, Daichi"
] | Holographic CCG Parsing | acl-long.15 | Oral | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.16.bib | https://aclanthology.org/2023.acl-long.16/ | @inproceedings{liang-etal-2023-prompts,
title = "Prompts Can Play Lottery Tickets Well: Achieving Lifelong Information Extraction via Lottery Prompt Tuning",
author = "Liang, Zujie and
Wei, Feng and
Jie, Yin and
Qian, Yuxi and
Hao, Zhenghong and
Han, Bing",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.16",
doi = "10.18653/v1/2023.acl-long.16",
pages = "277--292",
abstract = "Thanks to the recent success of Pre-trained Language Models (PLMs), it has become a promising research direction to develop a universal model (UIE) that can solve all typical information extraction tasks within one generative framework. Nonetheless, in real-world scenarios of UIE applications, new data of different IE tasks and domains usually come in a stream over time. A desirable UIE system should be capable of continually learning new tasks without forgetting old ones, thereby allowing knowledge and functionalities expansion without re-training the whole system. In this paper, we study the UIE system under a more challenging yet practical scenario, i.e., {``}lifelong learning{''} settings, to evaluate its abilities in three aspects, including knowledge sharing and expansion, catastrophic forgetting prevention, and rapid generalization on few-shot and unseen tasks. To achieve these three goals, we present a novel parameter- and deployment-efficient prompt tuning method namely Lottery Prompt Tuning (LPT).LPT freezes the PLM{'}s parameters and sequentially learns compact pruned prompt vectors for each task leveraging a binary prompt mask, while keeping the prompt parameters selected by the previous tasks insusceptible. Furthermore, we use a simple yet effective method to perform mask selection and show the powerful transferability of Lottery Prompts to novel tasks. Extensive experiments demonstrate that LPT consistently sets state-of-the-art performance on multiple lifelong learning settings of UIE, including task-incremental setting on seen tasks, few-shot adaptation, and zero-shot generalization on novel tasks.",
}
| Thanks to the recent success of Pre-trained Language Models (PLMs), it has become a promising research direction to develop a universal model (UIE) that can solve all typical information extraction tasks within one generative framework. Nonetheless, in real-world scenarios of UIE applications, new data of different IE tasks and domains usually come in a stream over time. A desirable UIE system should be capable of continually learning new tasks without forgetting old ones, thereby allowing knowledge and functionalities expansion without re-training the whole system. In this paper, we study the UIE system under a more challenging yet practical scenario, i.e., {``}lifelong learning{''} settings, to evaluate its abilities in three aspects, including knowledge sharing and expansion, catastrophic forgetting prevention, and rapid generalization on few-shot and unseen tasks. To achieve these three goals, we present a novel parameter- and deployment-efficient prompt tuning method namely Lottery Prompt Tuning (LPT).LPT freezes the PLM{'}s parameters and sequentially learns compact pruned prompt vectors for each task leveraging a binary prompt mask, while keeping the prompt parameters selected by the previous tasks insusceptible. Furthermore, we use a simple yet effective method to perform mask selection and show the powerful transferability of Lottery Prompts to novel tasks. Extensive experiments demonstrate that LPT consistently sets state-of-the-art performance on multiple lifelong learning settings of UIE, including task-incremental setting on seen tasks, few-shot adaptation, and zero-shot generalization on novel tasks. | [
"Liang, Zujie",
"Wei, Feng",
"Jie, Yin",
"Qian, Yuxi",
"Hao, Zhenghong",
"Han, Bing"
] | Prompts Can Play Lottery Tickets Well: Achieving Lifelong Information Extraction via Lottery Prompt Tuning | acl-long.16 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.17.bib | https://aclanthology.org/2023.acl-long.17/ | @inproceedings{ren-etal-2023-retrieve,
title = "Retrieve-and-Sample: Document-level Event Argument Extraction via Hybrid Retrieval Augmentation",
author = "Ren, Yubing and
Cao, Yanan and
Guo, Ping and
Fang, Fang and
Ma, Wei and
Lin, Zheng",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.17",
doi = "10.18653/v1/2023.acl-long.17",
pages = "293--306",
abstract = "Recent studies have shown the effectiveness of retrieval augmentation in many generative NLP tasks. These retrieval-augmented methods allow models to explicitly acquire prior external knowledge in a non-parametric manner and regard the retrieved reference instances as cues to augment text generation. These methods use similarity-based retrieval, which is based on a simple hypothesis: the more the retrieved demonstration resembles the original input, the more likely the demonstration label resembles the input label. However, due to the complexity of event labels and sparsity of event arguments, this hypothesis does not always hold in document-level EAE. This raises an interesting question: How do we design the retrieval strategy for document-level EAE? We investigate various retrieval settings from the input and label distribution views in this paper. We further augment document-level EAE with pseudo demonstrations sampled from event semantic regions that can cover adequate alternatives in the same context and event schema. Through extensive experiments on RAMS and WikiEvents, we demonstrate the validity of our newly introduced retrieval-augmented methods and analyze why they work.",
}
| Recent studies have shown the effectiveness of retrieval augmentation in many generative NLP tasks. These retrieval-augmented methods allow models to explicitly acquire prior external knowledge in a non-parametric manner and regard the retrieved reference instances as cues to augment text generation. These methods use similarity-based retrieval, which is based on a simple hypothesis: the more the retrieved demonstration resembles the original input, the more likely the demonstration label resembles the input label. However, due to the complexity of event labels and sparsity of event arguments, this hypothesis does not always hold in document-level EAE. This raises an interesting question: How do we design the retrieval strategy for document-level EAE? We investigate various retrieval settings from the input and label distribution views in this paper. We further augment document-level EAE with pseudo demonstrations sampled from event semantic regions that can cover adequate alternatives in the same context and event schema. Through extensive experiments on RAMS and WikiEvents, we demonstrate the validity of our newly introduced retrieval-augmented methods and analyze why they work. | [
"Ren, Yubing",
"Cao, Yanan",
"Guo, Ping",
"Fang, Fang",
"Ma, Wei",
"Lin, Zheng"
] | Retrieve-and-Sample: Document-level Event Argument Extraction via Hybrid Retrieval Augmentation | acl-long.17 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.18.bib | https://aclanthology.org/2023.acl-long.18/ | @inproceedings{wu-etal-2023-wecheck,
title = "{W}e{C}heck: Strong Factual Consistency Checker via Weakly Supervised Learning",
author = "Wu, Wenhao and
Li, Wei and
Xiao, Xinyan and
Liu, Jiachen and
Li, Sujian and
Lyu, Yajuan",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.18",
doi = "10.18653/v1/2023.acl-long.18",
pages = "307--321",
abstract = "A crucial issue of current text generation models is that they often uncontrollably generate text that is factually inconsistent with inputs. Due to lack of annotated data, existing factual consistency metrics usually train evaluation models on synthetic texts or directly transfer from other related tasks, such as question answering (QA) and natural language inference (NLI).Bias in synthetic text or upstream tasks makes them perform poorly on text actually generated by language models, especially for general evaluation for various tasks. To alleviate this problem, we propose a weakly supervised framework named \textbf{WeCheck} that is directly trained on actual generated samples from language models with weakly annotated labels.WeCheck first utilizes a generative model to infer the factual labels of generated samples by aggregating weak labels from multiple resources.Next, we train a simple noise-aware classification model as the target metric using the inferred weakly supervised information.Comprehensive experiments on various tasks demonstrate the strong performance of WeCheck, achieving an average absolute improvement of 3.3{\%} on the TRUE benchmark over 11B state-of-the-art methods using only 435M parameters.Furthermore, it is up to 30 times faster than previous evaluation methods, greatly improving the accuracy and efficiency of factual consistency evaluation.",
}
| A crucial issue of current text generation models is that they often uncontrollably generate text that is factually inconsistent with inputs. Due to lack of annotated data, existing factual consistency metrics usually train evaluation models on synthetic texts or directly transfer from other related tasks, such as question answering (QA) and natural language inference (NLI).Bias in synthetic text or upstream tasks makes them perform poorly on text actually generated by language models, especially for general evaluation for various tasks. To alleviate this problem, we propose a weakly supervised framework named \textbf{WeCheck} that is directly trained on actual generated samples from language models with weakly annotated labels.WeCheck first utilizes a generative model to infer the factual labels of generated samples by aggregating weak labels from multiple resources.Next, we train a simple noise-aware classification model as the target metric using the inferred weakly supervised information.Comprehensive experiments on various tasks demonstrate the strong performance of WeCheck, achieving an average absolute improvement of 3.3{\%} on the TRUE benchmark over 11B state-of-the-art methods using only 435M parameters.Furthermore, it is up to 30 times faster than previous evaluation methods, greatly improving the accuracy and efficiency of factual consistency evaluation. | [
"Wu, Wenhao",
"Li, Wei",
"Xiao, Xinyan",
"Liu, Jiachen",
"Li, Sujian",
"Lyu, Yajuan"
] | WeCheck: Strong Factual Consistency Checker via Weakly Supervised Learning | acl-long.18 | Poster | 2212.10057 | [
""
] | https://huggingface.co./papers/2212.10057 | 0 | 0 | 0 | 6 | 1 | [
"nightdessert/WeCheck"
] | [] | [] |
https://aclanthology.org/2023.acl-long.19.bib | https://aclanthology.org/2023.acl-long.19/ | @inproceedings{ma-etal-2023-amr,
title = "{AMR}-based Network for Aspect-based Sentiment Analysis",
author = "Ma, Fukun and
Hu, Xuming and
Liu, Aiwei and
Yang, Yawen and
Li, Shuang and
Yu, Philip S. and
Wen, Lijie",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.19",
doi = "10.18653/v1/2023.acl-long.19",
pages = "322--337",
abstract = "Aspect-based sentiment analysis (ABSA) is a fine-grained sentiment classification task. Many recent works have used dependency trees to extract the relation between aspects and contexts and have achieved significant improvements. However, further improvement is limited due to the potential mismatch between the dependency tree as a syntactic structure and the sentiment classification as a semantic task. To alleviate this gap, we replace the syntactic dependency tree with the semantic structure named Abstract Meaning Representation (AMR) and propose a model called AMR-based Path Aggregation Relational Network (APARN) to take full advantage of semantic structures. In particular, we design the path aggregator and the relation-enhanced self-attention mechanism that complement each other. The path aggregator extracts semantic features from AMRs under the guidance of sentence information, while the relation-enhanced self-attention mechanism in turn improves sentence features with refined semantic information. Experimental results on four public datasets demonstrate 1.13{\%} average F1 improvement of APARN in ABSA when compared with state-of-the-art baselines.",
}
| Aspect-based sentiment analysis (ABSA) is a fine-grained sentiment classification task. Many recent works have used dependency trees to extract the relation between aspects and contexts and have achieved significant improvements. However, further improvement is limited due to the potential mismatch between the dependency tree as a syntactic structure and the sentiment classification as a semantic task. To alleviate this gap, we replace the syntactic dependency tree with the semantic structure named Abstract Meaning Representation (AMR) and propose a model called AMR-based Path Aggregation Relational Network (APARN) to take full advantage of semantic structures. In particular, we design the path aggregator and the relation-enhanced self-attention mechanism that complement each other. The path aggregator extracts semantic features from AMRs under the guidance of sentence information, while the relation-enhanced self-attention mechanism in turn improves sentence features with refined semantic information. Experimental results on four public datasets demonstrate 1.13{\%} average F1 improvement of APARN in ABSA when compared with state-of-the-art baselines. | [
"Ma, Fukun",
"Hu, Xuming",
"Liu, Aiwei",
"Yang, Yawen",
"Li, Shuang",
"Yu, Philip S.",
"Wen, Lijie"
] | AMR-based Network for Aspect-based Sentiment Analysis | acl-long.19 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.20.bib | https://aclanthology.org/2023.acl-long.20/ | @inproceedings{li-etal-2023-text,
title = "Text Adversarial Purification as Defense against Adversarial Attacks",
author = "Li, Linyang and
Song, Demin and
Qiu, Xipeng",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.20",
doi = "10.18653/v1/2023.acl-long.20",
pages = "338--350",
abstract = "Adversarial purification is a successful defense mechanism against adversarial attacks without requiring knowledge of the form of the incoming attack. Generally, adversarial purification aims to remove the adversarial perturbations therefore can make correct predictions based on the recovered clean samples. Despite the success of adversarial purification in the computer vision field that incorporates generative models such as energy-based models and diffusion models,using purification as a defense strategy against textual adversarial attacks is rarely explored. In this work, we introduce a novel adversarial purification method that focuses on defending against textual adversarial attacks. With the help of language models, we can inject noise by masking input texts and reconstructing the masked texts based on the masked language models. In this way, we construct an adversarial purification process for textual models against the most widely used word-substitution adversarial attacks. We test our proposed adversarial purification method on several strong adversarial attack methods including Textfooler and BERT-Attack and experimental results indicate that the purification algorithm can successfully defend against strong word-substitution attacks.",
}
| Adversarial purification is a successful defense mechanism against adversarial attacks without requiring knowledge of the form of the incoming attack. Generally, adversarial purification aims to remove the adversarial perturbations therefore can make correct predictions based on the recovered clean samples. Despite the success of adversarial purification in the computer vision field that incorporates generative models such as energy-based models and diffusion models,using purification as a defense strategy against textual adversarial attacks is rarely explored. In this work, we introduce a novel adversarial purification method that focuses on defending against textual adversarial attacks. With the help of language models, we can inject noise by masking input texts and reconstructing the masked texts based on the masked language models. In this way, we construct an adversarial purification process for textual models against the most widely used word-substitution adversarial attacks. We test our proposed adversarial purification method on several strong adversarial attack methods including Textfooler and BERT-Attack and experimental results indicate that the purification algorithm can successfully defend against strong word-substitution attacks. | [
"Li, Linyang",
"Song, Demin",
"Qiu, Xipeng"
] | Text Adversarial Purification as Defense against Adversarial Attacks | acl-long.20 | Poster | 2203.14207 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.21.bib | https://aclanthology.org/2023.acl-long.21/ | @inproceedings{deng-etal-2023-speech,
title = "{SPEECH}: Structured Prediction with Energy-Based Event-Centric Hyperspheres",
author = "Deng, Shumin and
Mao, Shengyu and
Zhang, Ningyu and
Hooi, Bryan",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.21",
doi = "10.18653/v1/2023.acl-long.21",
pages = "351--363",
abstract = "Event-centric structured prediction involves predicting structured outputs of events. In most NLP cases, event structures are complex with manifold dependency, and it is challenging to effectively represent these complicated structured events. To address these issues, we propose Structured Prediction with Energy-based Event-Centric Hyperspheres (SPEECH). SPEECH models complex dependency among event structured components with energy-based modeling, and represents event classes with simple but effective hyperspheres. Experiments on two unified-annotated event datasets indicate that SPEECH is predominant in event detection and event-relation extraction tasks.",
}
| Event-centric structured prediction involves predicting structured outputs of events. In most NLP cases, event structures are complex with manifold dependency, and it is challenging to effectively represent these complicated structured events. To address these issues, we propose Structured Prediction with Energy-based Event-Centric Hyperspheres (SPEECH). SPEECH models complex dependency among event structured components with energy-based modeling, and represents event classes with simple but effective hyperspheres. Experiments on two unified-annotated event datasets indicate that SPEECH is predominant in event detection and event-relation extraction tasks. | [
"Deng, Shumin",
"Mao, Shengyu",
"Zhang, Ningyu",
"Hooi, Bryan"
] | SPEECH: Structured Prediction with Energy-Based Event-Centric Hyperspheres | acl-long.21 | Poster | 2305.13617 | [
"https://github.com/zjunlp/speech"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.22.bib | https://aclanthology.org/2023.acl-long.22/ | @inproceedings{clarke-etal-2023-rule,
title = "Rule By Example: Harnessing Logical Rules for Explainable Hate Speech Detection",
author = "Clarke, Christopher and
Hall, Matthew and
Mittal, Gaurav and
Yu, Ye and
Sajeev, Sandra and
Mars, Jason and
Chen, Mei",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.22",
doi = "10.18653/v1/2023.acl-long.22",
pages = "364--376",
abstract = "Classic approaches to content moderation typically apply a rule-based heuristic approach to flag content. While rules are easily customizable and intuitive for humans to interpret, they are inherently fragile and lack the flexibility or robustness needed to moderate the vast amount of undesirable content found online today. Recent advances in deep learning have demonstrated the promise of using highly effective deep neural models to overcome these challenges. However, despite the improved performance, these data-driven models lack transparency and explainability, often leading to mistrust from everyday users and a lack of adoption by many platforms. In this paper, we present Rule By Example (RBE): a novel exemplar-based contrastive learning approach for learning from logical rules for the task of textual content moderation. RBE is capable of providing rule-grounded predictions, allowing for more explainable and customizable predictions compared to typical deep learning-based approaches. We demonstrate that our approach is capable of learning rich rule embedding representations using only a few data examples. Experimental results on 3 popular hate speech classification datasets show that RBE is able to outperform state-of-the-art deep learning classifiers as well as the use of rules in both supervised and unsupervised settings while providing explainable model predictions via rule-grounding.",
}
| Classic approaches to content moderation typically apply a rule-based heuristic approach to flag content. While rules are easily customizable and intuitive for humans to interpret, they are inherently fragile and lack the flexibility or robustness needed to moderate the vast amount of undesirable content found online today. Recent advances in deep learning have demonstrated the promise of using highly effective deep neural models to overcome these challenges. However, despite the improved performance, these data-driven models lack transparency and explainability, often leading to mistrust from everyday users and a lack of adoption by many platforms. In this paper, we present Rule By Example (RBE): a novel exemplar-based contrastive learning approach for learning from logical rules for the task of textual content moderation. RBE is capable of providing rule-grounded predictions, allowing for more explainable and customizable predictions compared to typical deep learning-based approaches. We demonstrate that our approach is capable of learning rich rule embedding representations using only a few data examples. Experimental results on 3 popular hate speech classification datasets show that RBE is able to outperform state-of-the-art deep learning classifiers as well as the use of rules in both supervised and unsupervised settings while providing explainable model predictions via rule-grounding. | [
"Clarke, Christopher",
"Hall, Matthew",
"Mittal, Gaurav",
"Yu, Ye",
"Sajeev, S",
"ra",
"Mars, Jason",
"Chen, Mei"
] | Rule By Example: Harnessing Logical Rules for Explainable Hate Speech Detection | acl-long.22 | Poster | 2307.12935 | [
"https://github.com/chrisisking/rule-by-example"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.23.bib | https://aclanthology.org/2023.acl-long.23/ | @inproceedings{lauscher-etal-2023-em,
title = "What about {``}em{''}? How Commercial Machine Translation Fails to Handle (Neo-)Pronouns",
author = "Lauscher, Anne and
Nozza, Debora and
Miltersen, Ehm and
Crowley, Archie and
Hovy, Dirk",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.23",
doi = "10.18653/v1/2023.acl-long.23",
pages = "377--392",
abstract = "As 3rd-person pronoun usage shifts to include novel forms, e.g., neopronouns, we need more research on identity-inclusive NLP. Exclusion is particularly harmful in one of the most popular NLP applications, machine translation (MT). Wrong pronoun translations can discriminate against marginalized groups, e.g., non-binary individuals (Dev et al., 2021). In this {``}reality check{''}, we study how three commercial MT systems translate 3rd-person pronouns. Concretely, we compare the translations of gendered vs. gender-neutral pronouns from English to five other languages (Danish, Farsi, French, German, Italian), and vice versa, from Danish to English.Our error analysis shows that the presence of a gender-neutral pronoun often leads to grammatical and semantic translation errors. Similarly, gender neutrality is often not preserved. By surveying the opinions of affected native speakers from diverse languages, we provide recommendations to address the issue in future MT research.",
}
| As 3rd-person pronoun usage shifts to include novel forms, e.g., neopronouns, we need more research on identity-inclusive NLP. Exclusion is particularly harmful in one of the most popular NLP applications, machine translation (MT). Wrong pronoun translations can discriminate against marginalized groups, e.g., non-binary individuals (Dev et al., 2021). In this {``}reality check{''}, we study how three commercial MT systems translate 3rd-person pronouns. Concretely, we compare the translations of gendered vs. gender-neutral pronouns from English to five other languages (Danish, Farsi, French, German, Italian), and vice versa, from Danish to English.Our error analysis shows that the presence of a gender-neutral pronoun often leads to grammatical and semantic translation errors. Similarly, gender neutrality is often not preserved. By surveying the opinions of affected native speakers from diverse languages, we provide recommendations to address the issue in future MT research. | [
"Lauscher, Anne",
"Nozza, Debora",
"Miltersen, Ehm",
"Crowley, Archie",
"Hovy, Dirk"
] | What about “em”? How Commercial Machine Translation Fails to Handle (Neo-)Pronouns | acl-long.23 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.24.bib | https://aclanthology.org/2023.acl-long.24/ | @inproceedings{zhang-etal-2023-overlap,
title = "What Is Overlap Knowledge in Event Argument Extraction? {APE}: A Cross-datasets Transfer Learning Model for {EAE}",
author = "Zhang, Kaihang and
Shuang, Kai and
Yang, Xinyue and
Yao, Xuyang and
Guo, Jinyu",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.24",
doi = "10.18653/v1/2023.acl-long.24",
pages = "393--409",
abstract = "The EAE task extracts a structured event record from an event text. Most existing approaches train the EAE model on each dataset independently and ignore the overlap knowledge across datasets. However, insufficient event records in a single dataset often prevent the existing model from achieving better performance. In this paper, we clearly define the overlap knowledge across datasets and split the knowledge of the EAE task into overlap knowledge across datasets and specific knowledge of the target dataset. We propose APE model to learn the two parts of knowledge in two serial learning phases without causing catastrophic forgetting. In addition, we formulate both learning phases as conditional generation tasks and design Stressing Entity Type Prompt to close the gap between the two phases. The experiments show APE achieves new state-of-the-art with a large margin in the EAE task. When only ten records are available in the target dataset, our model dramatically outperforms the baseline model with average 27.27{\%} F1 gain.",
}
| The EAE task extracts a structured event record from an event text. Most existing approaches train the EAE model on each dataset independently and ignore the overlap knowledge across datasets. However, insufficient event records in a single dataset often prevent the existing model from achieving better performance. In this paper, we clearly define the overlap knowledge across datasets and split the knowledge of the EAE task into overlap knowledge across datasets and specific knowledge of the target dataset. We propose APE model to learn the two parts of knowledge in two serial learning phases without causing catastrophic forgetting. In addition, we formulate both learning phases as conditional generation tasks and design Stressing Entity Type Prompt to close the gap between the two phases. The experiments show APE achieves new state-of-the-art with a large margin in the EAE task. When only ten records are available in the target dataset, our model dramatically outperforms the baseline model with average 27.27{\%} F1 gain. | [
"Zhang, Kaihang",
"Shuang, Kai",
"Yang, Xinyue",
"Yao, Xuyang",
"Guo, Jinyu"
] | What Is Overlap Knowledge in Event Argument Extraction? APE: A Cross-datasets Transfer Learning Model for EAE | acl-long.24 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.25.bib | https://aclanthology.org/2023.acl-long.25/ | @inproceedings{yang-etal-2023-tailor,
title = "Tailor: A Soft-Prompt-Based Approach to Attribute-Based Controlled Text Generation",
author = "Yang, Kexin and
Liu, Dayiheng and
Lei, Wenqiang and
Yang, Baosong and
Xue, Mingfeng and
Chen, Boxing and
Xie, Jun",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.25",
doi = "10.18653/v1/2023.acl-long.25",
pages = "410--427",
abstract = "Attribute-based Controlled Text Generation (CTG) refers to generating sentences that satisfy desirable attributes (e.g., emotions and topics). Existing work usually utilize fine-tuning or resort to extra attribute classifiers, yet suffer from increases in storage and inference time. To address these concerns, we explore attribute-based CTG in a parameter-efficient manner. In short, the proposed Tailor represents each attribute as a pre-trained continuous vector i.e., single-attribute prompt), which guides the generation of a fixed pre-trained language model (PLM) to satisfy a pre-specified attribute. These prompts can be simply concatenated as a whole for multi-attribute CTG without any re-training. Nevertheless, this may raise problems of fluency downgrading and position sensitivity. To solve this, Tailor provides two solutions to enhance the combination. The former contains a multi-attribute prompt mask and a re-indexing position sequence to bridge the gap between the training (one single-attribute prompt for each task) and the testing stage (concatenating two prompts). The latter introduces a trainable prompt connector to further enhance the combinations. Experiments demonstrate that, only requiring 0.08{\%} extra training parameters of the GPT-2, Tailor can achieve effective and general improvements on eleven attribute-specific generation tasks.",
}
| Attribute-based Controlled Text Generation (CTG) refers to generating sentences that satisfy desirable attributes (e.g., emotions and topics). Existing work usually utilize fine-tuning or resort to extra attribute classifiers, yet suffer from increases in storage and inference time. To address these concerns, we explore attribute-based CTG in a parameter-efficient manner. In short, the proposed Tailor represents each attribute as a pre-trained continuous vector i.e., single-attribute prompt), which guides the generation of a fixed pre-trained language model (PLM) to satisfy a pre-specified attribute. These prompts can be simply concatenated as a whole for multi-attribute CTG without any re-training. Nevertheless, this may raise problems of fluency downgrading and position sensitivity. To solve this, Tailor provides two solutions to enhance the combination. The former contains a multi-attribute prompt mask and a re-indexing position sequence to bridge the gap between the training (one single-attribute prompt for each task) and the testing stage (concatenating two prompts). The latter introduces a trainable prompt connector to further enhance the combinations. Experiments demonstrate that, only requiring 0.08{\%} extra training parameters of the GPT-2, Tailor can achieve effective and general improvements on eleven attribute-specific generation tasks. | [
"Yang, Kexin",
"Liu, Dayiheng",
"Lei, Wenqiang",
"Yang, Baosong",
"Xue, Mingfeng",
"Chen, Boxing",
"Xie, Jun"
] | Tailor: A Soft-Prompt-Based Approach to Attribute-Based Controlled Text Generation | acl-long.25 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.26.bib | https://aclanthology.org/2023.acl-long.26/ | @inproceedings{ramezani-xu-2023-knowledge,
title = "Knowledge of cultural moral norms in large language models",
author = "Ramezani, Aida and
Xu, Yang",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.26",
doi = "10.18653/v1/2023.acl-long.26",
pages = "428--446",
abstract = "Moral norms vary across cultures. A recent line of work suggests that English large language models contain human-like moral biases, but these studies typically do not examine moral variation in a diverse cultural setting. We investigate the extent to which monolingual English language models contain knowledge about moral norms in different countries. We consider two levels of analysis: 1) whether language models capture fine-grained moral variation across countries over a variety of topics such as {``}homosexuality{''} and {``}divorce{''}; 2) whether language models capture cultural diversity and shared tendencies in which topics people around the globe tend to diverge or agree on in their moral judgment. We perform our analyses with two public datasets from the World Values Survey (across 55 countries) and PEW global surveys (across 40 countries) on morality. We find that pre-trained English language models predict empirical moral norms across countries worse than the English moral norms reported previously. However, fine-tuning language models on the survey data improves inference across countries at the expense of a less accurate estimate of the English moral norms. We discuss the relevance and challenges of incorporating cultural knowledge into the automated inference of moral norms.",
}
| Moral norms vary across cultures. A recent line of work suggests that English large language models contain human-like moral biases, but these studies typically do not examine moral variation in a diverse cultural setting. We investigate the extent to which monolingual English language models contain knowledge about moral norms in different countries. We consider two levels of analysis: 1) whether language models capture fine-grained moral variation across countries over a variety of topics such as {``}homosexuality{''} and {``}divorce{''}; 2) whether language models capture cultural diversity and shared tendencies in which topics people around the globe tend to diverge or agree on in their moral judgment. We perform our analyses with two public datasets from the World Values Survey (across 55 countries) and PEW global surveys (across 40 countries) on morality. We find that pre-trained English language models predict empirical moral norms across countries worse than the English moral norms reported previously. However, fine-tuning language models on the survey data improves inference across countries at the expense of a less accurate estimate of the English moral norms. We discuss the relevance and challenges of incorporating cultural knowledge into the automated inference of moral norms. | [
"Ramezani, Aida",
"Xu, Yang"
] | Knowledge of cultural moral norms in large language models | acl-long.26 | Poster | 2306.01857 | [
"https://github.com/aidaramezani/cultural_inference"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.27.bib | https://aclanthology.org/2023.acl-long.27/ | @inproceedings{ou-etal-2023-songs,
title = "Songs Across Borders: Singable and Controllable Neural Lyric Translation",
author = "Ou, Longshen and
Ma, Xichu and
Kan, Min-Yen and
Wang, Ye",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.27",
doi = "10.18653/v1/2023.acl-long.27",
pages = "447--467",
abstract = "The development of general-domain neural machine translation (NMT) methods has advanced significantly in recent years, but the lack of naturalness and musical constraints in the outputs makes them unable to produce singable lyric translations. This paper bridges the singability quality gap by formalizing lyric translation into a constrained translation problem, converting theoretical guidance and practical techniques from translatology literature to prompt-driven NMT approaches, exploring better adaptation methods, and instantiating them to an English-Chinese lyric translation system. Our model achieves 99.85{\%}, 99.00{\%}, and 95.52{\%} on length accuracy, rhyme accuracy, and word boundary recall. In our subjective evaluation, our model shows a 75{\%} relative enhancement on overall quality, compared against naive fine-tuning (Code available at \url{https://github.com/Sonata165/ControllableLyricTranslation}).",
}
| The development of general-domain neural machine translation (NMT) methods has advanced significantly in recent years, but the lack of naturalness and musical constraints in the outputs makes them unable to produce singable lyric translations. This paper bridges the singability quality gap by formalizing lyric translation into a constrained translation problem, converting theoretical guidance and practical techniques from translatology literature to prompt-driven NMT approaches, exploring better adaptation methods, and instantiating them to an English-Chinese lyric translation system. Our model achieves 99.85{\%}, 99.00{\%}, and 95.52{\%} on length accuracy, rhyme accuracy, and word boundary recall. In our subjective evaluation, our model shows a 75{\%} relative enhancement on overall quality, compared against naive fine-tuning (Code available at \url{https://github.com/Sonata165/ControllableLyricTranslation}). | [
"Ou, Longshen",
"Ma, Xichu",
"Kan, Min-Yen",
"Wang, Ye"
] | Songs Across Borders: Singable and Controllable Neural Lyric Translation | acl-long.27 | Poster | 2305.16816 | [
"https://github.com/sonata165/controllablelyrictranslation"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.28.bib | https://aclanthology.org/2023.acl-long.28/ | @inproceedings{yang-etal-2023-fantastic,
title = "Fantastic Expressions and Where to Find Them: {C}hinese Simile Generation with Multiple Constraints",
author = "Yang, Kexin and
Liu, Dayiheng and
Lei, Wenqiang and
Yang, Baosong and
Wei, Xiangpeng and
Liu, Zhengyuan and
Xie, Jun",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.28",
doi = "10.18653/v1/2023.acl-long.28",
pages = "468--486",
abstract = "Similes occur in the creative context of describing a concept (i.e., tenor) by making a literally false yet figuratively meaningful comparison to another (i.e., vehicle). Previous efforts form simile generation as a context-free generation task, focusing on simile-style transfer or writing a simile from a given prefix. However, generated texts under such settings might be undesirable, such as hardly meeting the simile definition (e.g., missing vehicle) or difficult to address certain preferences of content as humans wish (e.g., describe the color of apples through the simile). We believe that a simile could be more qualified and user-oriented if incorporated with pre-specified constraints. To this end, we introduce controllable simile generation (CSG), a new task that requires the model to generate a simile with multiple simile elements, e.g., context and vehicle. To facilitate this task, we present GraCe, including 61.3k simile-element annotated Chinese similes. Based on it, we propose a CSG model Similor to benchmark this task, including a vehicle retrieval module Scorer to obtain the explicable comparison for a given tenor in the vehicle-unknown situation. Both statistical and experimental analyses show that GraCe is of high quality beyond all other Chinese simile datasets, in terms of the number (8 vs. 3) of annotation elements, Is-Simile accuracy (98.9{\%} vs. 78.7{\%}), and increasing model-performance gains for both uncontrollable and controllable simile generation. Meanwhile, Similor can serve as a strong baseline for CSG, especially with Scorer, which beats model-based retrieval methods without any re-training.",
}
| Similes occur in the creative context of describing a concept (i.e., tenor) by making a literally false yet figuratively meaningful comparison to another (i.e., vehicle). Previous efforts form simile generation as a context-free generation task, focusing on simile-style transfer or writing a simile from a given prefix. However, generated texts under such settings might be undesirable, such as hardly meeting the simile definition (e.g., missing vehicle) or difficult to address certain preferences of content as humans wish (e.g., describe the color of apples through the simile). We believe that a simile could be more qualified and user-oriented if incorporated with pre-specified constraints. To this end, we introduce controllable simile generation (CSG), a new task that requires the model to generate a simile with multiple simile elements, e.g., context and vehicle. To facilitate this task, we present GraCe, including 61.3k simile-element annotated Chinese similes. Based on it, we propose a CSG model Similor to benchmark this task, including a vehicle retrieval module Scorer to obtain the explicable comparison for a given tenor in the vehicle-unknown situation. Both statistical and experimental analyses show that GraCe is of high quality beyond all other Chinese simile datasets, in terms of the number (8 vs. 3) of annotation elements, Is-Simile accuracy (98.9{\%} vs. 78.7{\%}), and increasing model-performance gains for both uncontrollable and controllable simile generation. Meanwhile, Similor can serve as a strong baseline for CSG, especially with Scorer, which beats model-based retrieval methods without any re-training. | [
"Yang, Kexin",
"Liu, Dayiheng",
"Lei, Wenqiang",
"Yang, Baosong",
"Wei, Xiangpeng",
"Liu, Zhengyuan",
"Xie, Jun"
] | Fantastic Expressions and Where to Find Them: Chinese Simile Generation with Multiple Constraints | acl-long.28 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.29.bib | https://aclanthology.org/2023.acl-long.29/ | @inproceedings{lei-etal-2023-revealing,
title = "Revealing Single Frame Bias for Video-and-Language Learning",
author = "Lei, Jie and
Berg, Tamara and
Bansal, Mohit",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.29",
doi = "10.18653/v1/2023.acl-long.29",
pages = "487--507",
abstract = "Training an effective video-and-language model intuitively requires multiple frames as model inputs. However, it is unclear whether using multiple frames is beneficial to downstream tasks, and if yes, whether the performance gain is worth the drastically-increased computation and memory costs resulting from using more frames. In this work, we explore single-frame models for video-and-language learning. On a diverse set of video-and-language tasks (including text-to-video retrieval and video question answering), we show the surprising result that, with large-scale pre-training and a proper frame ensemble strategy at inference time, a single-frame trained model that does not consider temporal information can achieve better performance than existing methods that use multiple frames for training. This result reveals the existence of a strong {``}static appearance bias{''} in popular video-and-language datasets. Therefore, to allow for a more comprehensive evaluation of video-and-language models, we propose two new retrieval tasks based on existing fine-grained action recognition datasets that encourage temporal modeling. Our code is available at \url{https://github.com/jayleicn/singularity}.",
}
| Training an effective video-and-language model intuitively requires multiple frames as model inputs. However, it is unclear whether using multiple frames is beneficial to downstream tasks, and if yes, whether the performance gain is worth the drastically-increased computation and memory costs resulting from using more frames. In this work, we explore single-frame models for video-and-language learning. On a diverse set of video-and-language tasks (including text-to-video retrieval and video question answering), we show the surprising result that, with large-scale pre-training and a proper frame ensemble strategy at inference time, a single-frame trained model that does not consider temporal information can achieve better performance than existing methods that use multiple frames for training. This result reveals the existence of a strong {``}static appearance bias{''} in popular video-and-language datasets. Therefore, to allow for a more comprehensive evaluation of video-and-language models, we propose two new retrieval tasks based on existing fine-grained action recognition datasets that encourage temporal modeling. Our code is available at \url{https://github.com/jayleicn/singularity}. | [
"Lei, Jie",
"Berg, Tamara",
"Bansal, Mohit"
] | Revealing Single Frame Bias for Video-and-Language Learning | acl-long.29 | Poster | 2206.03428 | [
"https://github.com/jayleicn/singularity"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.30.bib | https://aclanthology.org/2023.acl-long.30/ | @inproceedings{liu-etal-2023-learning,
title = "Learning with Partial Annotations for Event Detection",
author = "Liu, Jian and
Sui, Dianbo and
Liu, Kang and
Liu, Haoyan and
Zhao, Zhe",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.30",
doi = "10.18653/v1/2023.acl-long.30",
pages = "508--523",
abstract = "Event detection (ED) seeks to discover and classify event instances in plain texts. Previous methods for ED typically adopt supervised learning, requiring fully labeled and high-quality training data. However, in a real-world application, we may not obtain clean training data but only partially labeled one, which could substantially impede the learning process. In this work, we conduct a seminal study for learning with partial annotations for ED.We propose a new trigger localization formulation using contrastive learning to distinguish ground-truth triggers from contexts, showing a decent robustness for addressing partial annotation noise. Impressively, in an extreme scenario where more than 90{\%} of events are unlabeled, our approach achieves an F1 score of over 60{\%}.In addition, we re-annotate and make available two fully annotated subsets of ACE 2005 to serve as an unbiased benchmark for event detection. We hope our approach and data will inspire future studies on this vital yet understudied problem.",
}
| Event detection (ED) seeks to discover and classify event instances in plain texts. Previous methods for ED typically adopt supervised learning, requiring fully labeled and high-quality training data. However, in a real-world application, we may not obtain clean training data but only partially labeled one, which could substantially impede the learning process. In this work, we conduct a seminal study for learning with partial annotations for ED.We propose a new trigger localization formulation using contrastive learning to distinguish ground-truth triggers from contexts, showing a decent robustness for addressing partial annotation noise. Impressively, in an extreme scenario where more than 90{\%} of events are unlabeled, our approach achieves an F1 score of over 60{\%}.In addition, we re-annotate and make available two fully annotated subsets of ACE 2005 to serve as an unbiased benchmark for event detection. We hope our approach and data will inspire future studies on this vital yet understudied problem. | [
"Liu, Jian",
"Sui, Dianbo",
"Liu, Kang",
"Liu, Haoyan",
"Zhao, Zhe"
] | Learning with Partial Annotations for Event Detection | acl-long.30 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.31.bib | https://aclanthology.org/2023.acl-long.31/ | @inproceedings{ma-etal-2023-world,
title = "World-to-Words: Grounded Open Vocabulary Acquisition through Fast Mapping in Vision-Language Models",
author = "Ma, Ziqiao and
Pan, Jiayi and
Chai, Joyce",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.31",
doi = "10.18653/v1/2023.acl-long.31",
pages = "524--544",
abstract = "The ability to connect language units to their referents in the physical world, referred to as grounding, is crucial to learning and understanding grounded meanings of words. While humans demonstrate fast mapping in new word learning, it remains unclear whether modern vision-language models can truly represent language with their grounded meanings, and how grounding may further bootstrap new word learning. To this end, we introduce Grounded Open Vocabulary Acquisition (GOVA) to examine grounding and bootstrapping in open-world language learning. As an initial attempt, we propose World-to-Words (W2W), a novel visually-grounded language model by pre-training on image-text pairs highlighting grounding as an objective. Through extensive experiments and analysis, we demonstrate that W2W is a more coherent and fast grounded word learner, and that the grounding ability acquired during pre-training helps the model to learn unseen words more rapidly and robustly.",
}
| The ability to connect language units to their referents in the physical world, referred to as grounding, is crucial to learning and understanding grounded meanings of words. While humans demonstrate fast mapping in new word learning, it remains unclear whether modern vision-language models can truly represent language with their grounded meanings, and how grounding may further bootstrap new word learning. To this end, we introduce Grounded Open Vocabulary Acquisition (GOVA) to examine grounding and bootstrapping in open-world language learning. As an initial attempt, we propose World-to-Words (W2W), a novel visually-grounded language model by pre-training on image-text pairs highlighting grounding as an objective. Through extensive experiments and analysis, we demonstrate that W2W is a more coherent and fast grounded word learner, and that the grounding ability acquired during pre-training helps the model to learn unseen words more rapidly and robustly. | [
"Ma, Ziqiao",
"Pan, Jiayi",
"Chai, Joyce"
] | World-to-Words: Grounded Open Vocabulary Acquisition through Fast Mapping in Vision-Language Models | acl-long.31 | Poster | 2306.08685 | [
"https://github.com/sled-group/world-to-words"
] | https://huggingface.co./papers/2306.08685 | 1 | 1 | 0 | 3 | 1 | [
"sled-umich/OctoBERT"
] | [
"sled-umich/GOVA-flickr"
] | [] |
https://aclanthology.org/2023.acl-long.32.bib | https://aclanthology.org/2023.acl-long.32/ | @inproceedings{stolfo-etal-2023-causal,
title = "A Causal Framework to Quantify the Robustness of Mathematical Reasoning with Language Models",
author = "Stolfo, Alessandro and
Jin, Zhijing and
Shridhar, Kumar and
Schoelkopf, Bernhard and
Sachan, Mrinmaya",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.32",
doi = "10.18653/v1/2023.acl-long.32",
pages = "545--561",
abstract = "We have recently witnessed a number of impressive results on hard mathematical reasoning problems with language models. At the same time, the robustness of these models has also been called into question; recent works have shown that models can rely on shallow patterns in the problem description when generating a solution. Building on the idea of behavioral testing, we propose a novel framework, which pins down the causal effect of various factors in the input, e.g., the surface form of the problem text, the operands, and math operators on the output solution. By grounding the behavioral analysis in a causal graph describing an intuitive reasoning process, we study the behavior of language models in terms of robustness and sensitivity to direct interventions in the input space. We apply our framework on a test bed of math word problems. Our analysis shows that robustness does not appear to continuously improve as a function of size, but the GPT-3 Davinci models (175B) achieve a dramatic improvement in both robustness and sensitivity compared to all other GPT variants.",
}
| We have recently witnessed a number of impressive results on hard mathematical reasoning problems with language models. At the same time, the robustness of these models has also been called into question; recent works have shown that models can rely on shallow patterns in the problem description when generating a solution. Building on the idea of behavioral testing, we propose a novel framework, which pins down the causal effect of various factors in the input, e.g., the surface form of the problem text, the operands, and math operators on the output solution. By grounding the behavioral analysis in a causal graph describing an intuitive reasoning process, we study the behavior of language models in terms of robustness and sensitivity to direct interventions in the input space. We apply our framework on a test bed of math word problems. Our analysis shows that robustness does not appear to continuously improve as a function of size, but the GPT-3 Davinci models (175B) achieve a dramatic improvement in both robustness and sensitivity compared to all other GPT variants. | [
"Stolfo, Aless",
"ro",
"Jin, Zhijing",
"Shridhar, Kumar",
"Schoelkopf, Bernhard",
"Sachan, Mrinmaya"
] | A Causal Framework to Quantify the Robustness of Mathematical Reasoning with Language Models | acl-long.32 | Poster | 2210.12023 | [
"https://github.com/alestolfo/causal-math"
] | https://huggingface.co./papers/2210.12023 | 0 | 0 | 0 | 5 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.33.bib | https://aclanthology.org/2023.acl-long.33/ | @inproceedings{zhao-etal-2023-evaluating,
title = "Evaluating Open-Domain Dialogues in Latent Space with Next Sentence Prediction and Mutual Information",
author = "Zhao, Kun and
Yang, Bohao and
Lin, Chenghua and
Rong, Wenge and
Villavicencio, Aline and
Cui, Xiaohui",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.33",
doi = "10.18653/v1/2023.acl-long.33",
pages = "562--574",
abstract = "The long-standing one-to-many issue of the open-domain dialogues poses significant challenges for automatic evaluation methods, i.e., there may be multiple suitable responses which differ in semantics for a given conversational context. To tackle this challenge, we propose a novel learning-based automatic evaluation metric (CMN), which can robustly evaluate open-domain dialogues by augmenting Conditional Variational Autoencoders (CVAEs) with a Next Sentence Prediction (NSP) objective and employing Mutual Information (MI) to model the semantic similarity of text in the latent space. Experimental results on two open-domain dialogue datasets demonstrate the superiority of our method compared with a wide range of baselines, especially in handling responses which are distant to the {``}golden{''} reference responses in semantics.",
}
| The long-standing one-to-many issue of the open-domain dialogues poses significant challenges for automatic evaluation methods, i.e., there may be multiple suitable responses which differ in semantics for a given conversational context. To tackle this challenge, we propose a novel learning-based automatic evaluation metric (CMN), which can robustly evaluate open-domain dialogues by augmenting Conditional Variational Autoencoders (CVAEs) with a Next Sentence Prediction (NSP) objective and employing Mutual Information (MI) to model the semantic similarity of text in the latent space. Experimental results on two open-domain dialogue datasets demonstrate the superiority of our method compared with a wide range of baselines, especially in handling responses which are distant to the {``}golden{''} reference responses in semantics. | [
"Zhao, Kun",
"Yang, Bohao",
"Lin, Chenghua",
"Rong, Wenge",
"Villavicencio, Aline",
"Cui, Xiaohui"
] | Evaluating Open-Domain Dialogues in Latent Space with Next Sentence Prediction and Mutual Information | acl-long.33 | Oral | 2305.16967 | [
"https://github.com/bernard-yang/cmn-acl2023"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.34.bib | https://aclanthology.org/2023.acl-long.34/ | @inproceedings{chung-etal-2023-increasing,
title = "Increasing Diversity While Maintaining Accuracy: Text Data Generation with Large Language Models and Human Interventions",
author = "Chung, John and
Kamar, Ece and
Amershi, Saleema",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.34",
doi = "10.18653/v1/2023.acl-long.34",
pages = "575--593",
abstract = "Large language models (LLMs) can be used to generate text data for training and evaluating other models. However, creating high-quality datasets with LLMs can be challenging. In this work, we explore human-AI partnerships to facilitate high diversity and accuracy in LLM-based text data generation. We first examine two approaches to diversify text generation: 1) logit suppression, which minimizes the generation of languages that have already been frequently generated, and 2) temperature sampling, which flattens the token sampling probability. We found that diversification approaches can increase data diversity but often at the cost of data accuracy (i.e., text and labels being appropriate for the target domain). To address this issue, we examined two human interventions, 1) label replacement (LR), correcting misaligned labels, and 2) out-of-scope filtering (OOSF), removing instances that are out of the user{'}s domain of interest or to which no considered label applies. With oracle studies, we found that LR increases the absolute accuracy of models trained with diversified datasets by 14.4{\%}. Moreover, we found that some models trained with data generated with LR interventions outperformed LLM-based few-shot classification. In contrast, OOSF was not effective in increasing model accuracy, implying the need for future work in human-in-the-loop text data generation.",
}
| Large language models (LLMs) can be used to generate text data for training and evaluating other models. However, creating high-quality datasets with LLMs can be challenging. In this work, we explore human-AI partnerships to facilitate high diversity and accuracy in LLM-based text data generation. We first examine two approaches to diversify text generation: 1) logit suppression, which minimizes the generation of languages that have already been frequently generated, and 2) temperature sampling, which flattens the token sampling probability. We found that diversification approaches can increase data diversity but often at the cost of data accuracy (i.e., text and labels being appropriate for the target domain). To address this issue, we examined two human interventions, 1) label replacement (LR), correcting misaligned labels, and 2) out-of-scope filtering (OOSF), removing instances that are out of the user{'}s domain of interest or to which no considered label applies. With oracle studies, we found that LR increases the absolute accuracy of models trained with diversified datasets by 14.4{\%}. Moreover, we found that some models trained with data generated with LR interventions outperformed LLM-based few-shot classification. In contrast, OOSF was not effective in increasing model accuracy, implying the need for future work in human-in-the-loop text data generation. | [
"Chung, John",
"Kamar, Ece",
"Amershi, Saleema"
] | Increasing Diversity While Maintaining Accuracy: Text Data Generation with Large Language Models and Human Interventions | acl-long.34 | Oral | 2306.04140 | [
""
] | https://huggingface.co./papers/2306.04140 | 2 | 2 | 0 | 3 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.35.bib | https://aclanthology.org/2023.acl-long.35/ | @inproceedings{jiang-etal-2023-pruning,
title = "Pruning Pre-trained Language Models Without Fine-Tuning",
author = "Jiang, Ting and
Wang, Deqing and
Zhuang, Fuzhen and
Xie, Ruobing and
Xia, Feng",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.35",
doi = "10.18653/v1/2023.acl-long.35",
pages = "594--605",
abstract = "To overcome the overparameterized problem in Pre-trained Language Models (PLMs), pruning is widely used as a simple and straightforward compression method by directly removing unimportant weights. Previous first-order methods successfully compress PLMs to extremely high sparsity with little performance drop. These methods, such as movement pruning, use first-order information to prune PLMs while fine-tuning the remaining weights. In this work, we argue fine-tuning is redundant for first-order pruning, since first-order pruning is sufficient to converge PLMs to downstream tasks without fine-tuning. Under this motivation, we propose Static Model Pruning (SMP), which only uses first-order pruning to adapt PLMs to downstream tasks while achieving the target sparsity level. In addition, we also design a new masking function and training objective to further improve SMP. Extensive experiments at various sparsity levels show SMP has significant improvements over first-order and zero-order methods. Unlike previous first-order methods, SMP is also applicable to low sparsity and outperforms zero-order methods. Meanwhile, SMP is more parameter efficient than other methods due to it does not require fine-tuning.",
}
| To overcome the overparameterized problem in Pre-trained Language Models (PLMs), pruning is widely used as a simple and straightforward compression method by directly removing unimportant weights. Previous first-order methods successfully compress PLMs to extremely high sparsity with little performance drop. These methods, such as movement pruning, use first-order information to prune PLMs while fine-tuning the remaining weights. In this work, we argue fine-tuning is redundant for first-order pruning, since first-order pruning is sufficient to converge PLMs to downstream tasks without fine-tuning. Under this motivation, we propose Static Model Pruning (SMP), which only uses first-order pruning to adapt PLMs to downstream tasks while achieving the target sparsity level. In addition, we also design a new masking function and training objective to further improve SMP. Extensive experiments at various sparsity levels show SMP has significant improvements over first-order and zero-order methods. Unlike previous first-order methods, SMP is also applicable to low sparsity and outperforms zero-order methods. Meanwhile, SMP is more parameter efficient than other methods due to it does not require fine-tuning. | [
"Jiang, Ting",
"Wang, Deqing",
"Zhuang, Fuzhen",
"Xie, Ruobing",
"Xia, Feng"
] | Pruning Pre-trained Language Models Without Fine-Tuning | acl-long.35 | Oral | 2210.06210 | [
"https://github.com/kongds/smp"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.36.bib | https://aclanthology.org/2023.acl-long.36/ | @inproceedings{fernandes-etal-2023-translation,
title = "When Does Translation Require Context? A Data-driven, Multilingual Exploration",
author = "Fernandes, Patrick and
Yin, Kayo and
Liu, Emmy and
Martins, Andr{\'e} and
Neubig, Graham",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.36",
doi = "10.18653/v1/2023.acl-long.36",
pages = "606--626",
abstract = "Although proper handling of discourse significantly contributes to the quality of machine translation (MT), these improvements are not adequately measured in common translation quality metrics. Recent works in context-aware MT attempt to target a small set of discourse phenomena during evaluation, however not in a fully systematic way. In this paper, we develop the Multilingual Discourse-Aware (MuDA) benchmark, a series of taggers that identify and evaluate model performance on discourse phenomena in any given dataset. The choice of phenomena is inspired by a novel methodology to systematically identify translations that require context. This methodology confirms the difficulty of previously studied phenomena while uncovering others which were not previously addressed. We find that commonly studied context-aware MT models make only marginal improvements over context-agnostic models, which suggests these models do not handle these ambiguities effectively. We release code and data for 14 language pairs to encourage the MT community to focus on accurately capturing discourse phenomena. Code available at \url{https://github.com/neulab/contextual-mt}",
}
| Although proper handling of discourse significantly contributes to the quality of machine translation (MT), these improvements are not adequately measured in common translation quality metrics. Recent works in context-aware MT attempt to target a small set of discourse phenomena during evaluation, however not in a fully systematic way. In this paper, we develop the Multilingual Discourse-Aware (MuDA) benchmark, a series of taggers that identify and evaluate model performance on discourse phenomena in any given dataset. The choice of phenomena is inspired by a novel methodology to systematically identify translations that require context. This methodology confirms the difficulty of previously studied phenomena while uncovering others which were not previously addressed. We find that commonly studied context-aware MT models make only marginal improvements over context-agnostic models, which suggests these models do not handle these ambiguities effectively. We release code and data for 14 language pairs to encourage the MT community to focus on accurately capturing discourse phenomena. Code available at \url{https://github.com/neulab/contextual-mt} | [
"Fern",
"es, Patrick",
"Yin, Kayo",
"Liu, Emmy",
"Martins, Andr{\\'e}",
"Neubig, Graham"
] | When Does Translation Require Context? A Data-driven, Multilingual Exploration | acl-long.36 | Poster | 2109.07446 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.37.bib | https://aclanthology.org/2023.acl-long.37/ | @inproceedings{chen-etal-2023-causal,
title = "Causal Intervention and Counterfactual Reasoning for Multi-modal Fake News Detection",
author = "Chen, Ziwei and
Hu, Linmei and
Li, Weixin and
Shao, Yingxia and
Nie, Liqiang",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.37",
doi = "10.18653/v1/2023.acl-long.37",
pages = "627--638",
abstract = "Due to the rapid upgrade of social platforms, most of today{'}s fake news is published and spread in a multi-modal form. Most existing multi-modal fake news detection methods neglect the fact that some label-specific features learned from the training set cannot generalize well to the testing set, thus inevitably suffering from the harm caused by the latent data bias. In this paper, we analyze and identify the psycholinguistic bias in the text and the bias of inferring news label based on only image features. We mitigate these biases from a causality perspective and propose a Causal intervention and Counterfactual reasoning based Debiasing framework (CCD) for multi-modal fake news detection. To achieve our goal, we first utilize causal intervention to remove the psycholinguistic bias which introduces the spurious correlations between text features and news label. And then, we apply counterfactual reasoning by imagining a counterfactual world where each news has only image features for estimating the direct effect of the image. Therefore we can eliminate the image-only bias by deducting the direct effect of the image from the total effect on labels. Extensive experiments on two real-world benchmark datasets demonstrate the effectiveness of our framework for improving multi-modal fake news detection.",
}
| Due to the rapid upgrade of social platforms, most of today{'}s fake news is published and spread in a multi-modal form. Most existing multi-modal fake news detection methods neglect the fact that some label-specific features learned from the training set cannot generalize well to the testing set, thus inevitably suffering from the harm caused by the latent data bias. In this paper, we analyze and identify the psycholinguistic bias in the text and the bias of inferring news label based on only image features. We mitigate these biases from a causality perspective and propose a Causal intervention and Counterfactual reasoning based Debiasing framework (CCD) for multi-modal fake news detection. To achieve our goal, we first utilize causal intervention to remove the psycholinguistic bias which introduces the spurious correlations between text features and news label. And then, we apply counterfactual reasoning by imagining a counterfactual world where each news has only image features for estimating the direct effect of the image. Therefore we can eliminate the image-only bias by deducting the direct effect of the image from the total effect on labels. Extensive experiments on two real-world benchmark datasets demonstrate the effectiveness of our framework for improving multi-modal fake news detection. | [
"Chen, Ziwei",
"Hu, Linmei",
"Li, Weixin",
"Shao, Yingxia",
"Nie, Liqiang"
] | Causal Intervention and Counterfactual Reasoning for Multi-modal Fake News Detection | acl-long.37 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.38.bib | https://aclanthology.org/2023.acl-long.38/ | @inproceedings{akyurek-andreas-2023-lexsym,
title = "{L}ex{S}ym: Compositionality as Lexical Symmetry",
author = "Akyurek, Ekin and
Andreas, Jacob",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.38",
doi = "10.18653/v1/2023.acl-long.38",
pages = "639--657",
abstract = "In tasks like semantic parsing, instruction following, and question answering, standard deep networks fail to generalize compositionally from small datasets. Many existing approaches overcome this limitation with model architectures that enforce a compositional process of sentence interpretation. In this paper, we present a domain-general and model-agnostic formulation of compositionality as a constraint on symmetries of data distributions rather than models. Informally, we prove that whenever a task can be solved by a compositional model, there is a corresponding data augmentation scheme {---} a procedure for transforming examples into other well-formed examples {---} that imparts compositional inductive bias on any model trained to solve the same task. We describe a procedure called LexSym that discovers these transformations automatically, then applies them to training data for ordinary neural sequence models. Unlike existing compositional data augmentation procedures, LexSym can be deployed agnostically across text, structured data, and even images. It matches or surpasses state-of-the-art, task-specific models on COGS semantic parsing, SCAN and Alchemy instruction following, and CLEVR-CoGenT visual question answering datasets.",
}
| In tasks like semantic parsing, instruction following, and question answering, standard deep networks fail to generalize compositionally from small datasets. Many existing approaches overcome this limitation with model architectures that enforce a compositional process of sentence interpretation. In this paper, we present a domain-general and model-agnostic formulation of compositionality as a constraint on symmetries of data distributions rather than models. Informally, we prove that whenever a task can be solved by a compositional model, there is a corresponding data augmentation scheme {---} a procedure for transforming examples into other well-formed examples {---} that imparts compositional inductive bias on any model trained to solve the same task. We describe a procedure called LexSym that discovers these transformations automatically, then applies them to training data for ordinary neural sequence models. Unlike existing compositional data augmentation procedures, LexSym can be deployed agnostically across text, structured data, and even images. It matches or surpasses state-of-the-art, task-specific models on COGS semantic parsing, SCAN and Alchemy instruction following, and CLEVR-CoGenT visual question answering datasets. | [
"Akyurek, Ekin",
"Andreas, Jacob"
] | LexSym: Compositionality as Lexical Symmetry | acl-long.38 | Oral | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.39.bib | https://aclanthology.org/2023.acl-long.39/ | @inproceedings{sun-etal-2023-layer,
title = "Layer-wise Fusion with Modality Independence Modeling for Multi-modal Emotion Recognition",
author = "Sun, Jun and
Han, Shoukang and
Ruan, Yu-Ping and
Zhang, Xiaoning and
Zheng, Shu-Kai and
Liu, Yulong and
Huang, Yuxin and
Li, Taihao",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.39",
doi = "10.18653/v1/2023.acl-long.39",
pages = "658--670",
abstract = "Multi-modal emotion recognition has gained increasing attention in recent years due to its widespread applications and the advances in multi-modal learning approaches. However, previous studies primarily focus on developing models that exploit the unification of multiple modalities. In this paper, we propose that maintaining modality independence is beneficial for the model performance. According to this principle, we construct a dataset, and devise a multi-modal transformer model. The new dataset, CHinese Emotion Recognition dataset with Modality-wise Annotions, abbreviated as CHERMA, provides uni-modal labels for each individual modality, and multi-modal labels for all modalities jointly observed. The model consists of uni-modal transformer modules that learn representations for each modality, and a multi-modal transformer module that fuses all modalities. All the modules are supervised by their corresponding labels separately, and the forward information flow is uni-directionally from the uni-modal modules to the multi-modal module. The supervision strategy and the model architecture guarantee each individual modality learns its representation independently, and meanwhile the multi-modal module aggregates all information. Extensive empirical results demonstrate that our proposed scheme outperforms state-of-the-art alternatives, corroborating the importance of modality independence in multi-modal emotion recognition. The dataset and codes are availabel at \url{https://github.com/sunjunaimer/LFMIM}",
}
| Multi-modal emotion recognition has gained increasing attention in recent years due to its widespread applications and the advances in multi-modal learning approaches. However, previous studies primarily focus on developing models that exploit the unification of multiple modalities. In this paper, we propose that maintaining modality independence is beneficial for the model performance. According to this principle, we construct a dataset, and devise a multi-modal transformer model. The new dataset, CHinese Emotion Recognition dataset with Modality-wise Annotions, abbreviated as CHERMA, provides uni-modal labels for each individual modality, and multi-modal labels for all modalities jointly observed. The model consists of uni-modal transformer modules that learn representations for each modality, and a multi-modal transformer module that fuses all modalities. All the modules are supervised by their corresponding labels separately, and the forward information flow is uni-directionally from the uni-modal modules to the multi-modal module. The supervision strategy and the model architecture guarantee each individual modality learns its representation independently, and meanwhile the multi-modal module aggregates all information. Extensive empirical results demonstrate that our proposed scheme outperforms state-of-the-art alternatives, corroborating the importance of modality independence in multi-modal emotion recognition. The dataset and codes are availabel at \url{https://github.com/sunjunaimer/LFMIM} | [
"Sun, Jun",
"Han, Shoukang",
"Ruan, Yu-Ping",
"Zhang, Xiaoning",
"Zheng, Shu-Kai",
"Liu, Yulong",
"Huang, Yuxin",
"Li, Taihao"
] | Layer-wise Fusion with Modality Independence Modeling for Multi-modal Emotion Recognition | acl-long.39 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.40.bib | https://aclanthology.org/2023.acl-long.40/ | @inproceedings{bao-etal-2023-casn,
title = "{CASN}:Class-Aware Score Network for Textual Adversarial Detection",
author = "Bao, Rong and
Zheng, Rui and
Ding, Liang and
Zhang, Qi and
Tao, Dacheng",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.40",
doi = "10.18653/v1/2023.acl-long.40",
pages = "671--687",
abstract = "Adversarial detection aims to detect adversarial samples that threaten the security of deep neural networks, which is an essential step toward building robust AI systems. Density-based estimation is widely considered as an effective technique by explicitly modeling the distribution of normal data and identifying adversarial ones as outliers. However, these methods suffer from significant performance degradation when the adversarial samples lie close to the non-adversarial data manifold. To address this limitation, we propose a score-based generative method to implicitly model the data distribution. Our approach utilizes the gradient of the log-density data distribution and calculates the distribution gap between adversarial and normal samples through multi-step iterations using Langevin dynamics. In addition, we use supervised contrastive learning to guide the gradient estimation using label information, which avoids collapsing to a single data manifold and better preserves the anisotropy of the different labeled data distributions. Experimental results on three text classification tasks upon four advanced attack algorithms show that our approach is a significant improvement (average +15.2 F1 score against previous SOTA) over previous detection methods.",
}
| Adversarial detection aims to detect adversarial samples that threaten the security of deep neural networks, which is an essential step toward building robust AI systems. Density-based estimation is widely considered as an effective technique by explicitly modeling the distribution of normal data and identifying adversarial ones as outliers. However, these methods suffer from significant performance degradation when the adversarial samples lie close to the non-adversarial data manifold. To address this limitation, we propose a score-based generative method to implicitly model the data distribution. Our approach utilizes the gradient of the log-density data distribution and calculates the distribution gap between adversarial and normal samples through multi-step iterations using Langevin dynamics. In addition, we use supervised contrastive learning to guide the gradient estimation using label information, which avoids collapsing to a single data manifold and better preserves the anisotropy of the different labeled data distributions. Experimental results on three text classification tasks upon four advanced attack algorithms show that our approach is a significant improvement (average +15.2 F1 score against previous SOTA) over previous detection methods. | [
"Bao, Rong",
"Zheng, Rui",
"Ding, Liang",
"Zhang, Qi",
"Tao, Dacheng"
] | CASN:Class-Aware Score Network for Textual Adversarial Detection | acl-long.40 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.41.bib | https://aclanthology.org/2023.acl-long.41/ | @inproceedings{hessel-etal-2023-androids,
title = "Do Androids Laugh at Electric Sheep? Humor {``}Understanding{''} Benchmarks from The New Yorker Caption Contest",
author = "Hessel, Jack and
Marasovic, Ana and
Hwang, Jena D. and
Lee, Lillian and
Da, Jeff and
Zellers, Rowan and
Mankoff, Robert and
Choi, Yejin",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.41",
doi = "10.18653/v1/2023.acl-long.41",
pages = "688--714",
abstract = "Large neural networks can now generate jokes, but do they really {``}understand{''} humor? We challenge AI models with three tasks derived from the New Yorker Cartoon Caption Contest: matching a joke to a cartoon, identifying a winning caption, and explaining why a winning caption is funny. These tasks encapsulate progressively more sophisticated aspects of {``}understanding{''} a cartoon; key elements are the complex, often surprising relationships between images and captions and the frequent inclusion of indirect and playful allusions to human experience and culture. We investigate both multimodal and language-only models: the former are challenged with the cartoon images directly, while the latter are given multifaceted descriptions of the visual scene to simulate human-level visual understanding. We find that both types of models struggle at all three tasks. For example, our best multimodal models fall 30 accuracy points behind human performance on the matching task, and, even when provided ground-truth visual scene descriptors, human-authored explanations are preferred head-to-head over the best machine-authored ones (few-shot GPT-4) in more than 2/3 of cases. We release models, code, leaderboard, and corpus, which includes newly-gathered annotations describing the image{'}s locations/entities, what{'}s unusual in the scene, and an explanation of the joke.",
}
| Large neural networks can now generate jokes, but do they really {``}understand{''} humor? We challenge AI models with three tasks derived from the New Yorker Cartoon Caption Contest: matching a joke to a cartoon, identifying a winning caption, and explaining why a winning caption is funny. These tasks encapsulate progressively more sophisticated aspects of {``}understanding{''} a cartoon; key elements are the complex, often surprising relationships between images and captions and the frequent inclusion of indirect and playful allusions to human experience and culture. We investigate both multimodal and language-only models: the former are challenged with the cartoon images directly, while the latter are given multifaceted descriptions of the visual scene to simulate human-level visual understanding. We find that both types of models struggle at all three tasks. For example, our best multimodal models fall 30 accuracy points behind human performance on the matching task, and, even when provided ground-truth visual scene descriptors, human-authored explanations are preferred head-to-head over the best machine-authored ones (few-shot GPT-4) in more than 2/3 of cases. We release models, code, leaderboard, and corpus, which includes newly-gathered annotations describing the image{'}s locations/entities, what{'}s unusual in the scene, and an explanation of the joke. | [
"Hessel, Jack",
"Marasovic, Ana",
"Hwang, Jena D.",
"Lee, Lillian",
"Da, Jeff",
"Zellers, Rowan",
"Mankoff, Robert",
"Choi, Yejin"
] | Do Androids Laugh at Electric Sheep? Humor “Understanding” Benchmarks from The New Yorker Caption Contest | acl-long.41 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.42.bib | https://aclanthology.org/2023.acl-long.42/ | @inproceedings{bartelds-etal-2023-making,
title = "Making More of Little Data: Improving Low-Resource Automatic Speech Recognition Using Data Augmentation",
author = "Bartelds, Martijn and
San, Nay and
McDonnell, Bradley and
Jurafsky, Dan and
Wieling, Martijn",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.42",
doi = "10.18653/v1/2023.acl-long.42",
pages = "715--729",
abstract = "The performance of automatic speech recognition (ASR) systems has advanced substantially in recent years, particularly for languages for which a large amount of transcribed speech is available. Unfortunately, for low-resource languages, such as minority languages, regional languages or dialects, ASR performance generally remains much lower. In this study, we investigate whether data augmentation techniques could help improve low-resource ASR performance, focusing on four typologically diverse minority languages or language variants (West Germanic: Gronings, West-Frisian; Malayo-Polynesian: Besemah, Nasal). For all four languages, we examine the use of self-training, where an ASR system trained with the available human-transcribed data is used to generate transcriptions, which are then combined with the original data to train a new ASR system. For Gronings, for which there was a pre-existing text-to-speech (TTS) system available, we also examined the use of TTS to generate ASR training data from text-only sources. We find that using a self-training approach consistently yields improved performance (a relative WER reduction up to 20.5{\%} compared to using an ASR system trained on 24 minutes of manually transcribed speech). The performance gain from TTS augmentation for Gronings was even stronger (up to 25.5{\%} relative reduction in WER compared to a system based on 24 minutes of manually transcribed speech). In sum, our results show the benefit of using self-training or (if possible) TTS-generated data as an efficient solution to overcome the limitations of data availability for resource-scarce languages in order to improve ASR performance.",
}
| The performance of automatic speech recognition (ASR) systems has advanced substantially in recent years, particularly for languages for which a large amount of transcribed speech is available. Unfortunately, for low-resource languages, such as minority languages, regional languages or dialects, ASR performance generally remains much lower. In this study, we investigate whether data augmentation techniques could help improve low-resource ASR performance, focusing on four typologically diverse minority languages or language variants (West Germanic: Gronings, West-Frisian; Malayo-Polynesian: Besemah, Nasal). For all four languages, we examine the use of self-training, where an ASR system trained with the available human-transcribed data is used to generate transcriptions, which are then combined with the original data to train a new ASR system. For Gronings, for which there was a pre-existing text-to-speech (TTS) system available, we also examined the use of TTS to generate ASR training data from text-only sources. We find that using a self-training approach consistently yields improved performance (a relative WER reduction up to 20.5{\%} compared to using an ASR system trained on 24 minutes of manually transcribed speech). The performance gain from TTS augmentation for Gronings was even stronger (up to 25.5{\%} relative reduction in WER compared to a system based on 24 minutes of manually transcribed speech). In sum, our results show the benefit of using self-training or (if possible) TTS-generated data as an efficient solution to overcome the limitations of data availability for resource-scarce languages in order to improve ASR performance. | [
"Bartelds, Martijn",
"San, Nay",
"McDonnell, Bradley",
"Jurafsky, Dan",
"Wieling, Martijn"
] | Making More of Little Data: Improving Low-Resource Automatic Speech Recognition Using Data Augmentation | acl-long.42 | Oral | 2305.10951 | [
"https://github.com/bartelds/asr-augmentation"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.43.bib | https://aclanthology.org/2023.acl-long.43/ | @inproceedings{zhou-etal-2023-clcl,
title = "{CLCL}: Non-compositional Expression Detection with Contrastive Learning and Curriculum Learning",
author = "Zhou, Jianing and
Zeng, Ziheng and
Bhat, Suma",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.43",
doi = "10.18653/v1/2023.acl-long.43",
pages = "730--743",
abstract = "Non-compositional expressions present a substantial challenge for natural language processing (NLP) systems, necessitating more intricate processing compared to general language tasks, even with large pre-trained language models. Their non-compositional nature and limited availability of data resources further compound the difficulties in accurately learning their representations. This paper addresses both of these challenges. By leveraging contrastive learning techniques to build improved representations it tackles the non-compositionality challenge. Additionally, we propose a dynamic curriculum learning framework specifically designed to take advantage of the scarce available data for modeling non-compositionality. Our framework employs an easy-to-hard learning strategy, progressively optimizing the model{'}s performance by effectively utilizing available training data. Moreover, we integrate contrastive learning into the curriculum learning approach to maximize its benefits. Experimental results demonstrate the gradual improvement in the model{'}s performance on idiom usage recognition and metaphor detection tasks. Our evaluation encompasses six datasets, consistently affirming the effectiveness of the proposed framework. Our models available at \url{https://github.com/zhjjn/CLCL.git}.",
}
| Non-compositional expressions present a substantial challenge for natural language processing (NLP) systems, necessitating more intricate processing compared to general language tasks, even with large pre-trained language models. Their non-compositional nature and limited availability of data resources further compound the difficulties in accurately learning their representations. This paper addresses both of these challenges. By leveraging contrastive learning techniques to build improved representations it tackles the non-compositionality challenge. Additionally, we propose a dynamic curriculum learning framework specifically designed to take advantage of the scarce available data for modeling non-compositionality. Our framework employs an easy-to-hard learning strategy, progressively optimizing the model{'}s performance by effectively utilizing available training data. Moreover, we integrate contrastive learning into the curriculum learning approach to maximize its benefits. Experimental results demonstrate the gradual improvement in the model{'}s performance on idiom usage recognition and metaphor detection tasks. Our evaluation encompasses six datasets, consistently affirming the effectiveness of the proposed framework. Our models available at \url{https://github.com/zhjjn/CLCL.git}. | [
"Zhou, Jianing",
"Zeng, Ziheng",
"Bhat, Suma"
] | CLCL: Non-compositional Expression Detection with Contrastive Learning and Curriculum Learning | acl-long.43 | Oral | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.44.bib | https://aclanthology.org/2023.acl-long.44/ | @inproceedings{ziems-etal-2023-multi,
title = "Multi-{VALUE}: A Framework for Cross-Dialectal {E}nglish {NLP}",
author = "Ziems, Caleb and
Held, William and
Yang, Jingfeng and
Dhamala, Jwala and
Gupta, Rahul and
Yang, Diyi",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.44",
doi = "10.18653/v1/2023.acl-long.44",
pages = "744--768",
abstract = "Dialect differences caused by regional, social, and economic factors cause performance discrepancies for many groups of language technology users. Inclusive and equitable language technology must critically be dialect invariant, meaning that performance remains constant over dialectal shifts. Current systems often fall short of this ideal since they are designed and tested on a single dialect: Standard American English (SAE). We introduce a suite of resources for evaluating and achieving English dialect invariance. The resource is called Multi-VALUE, a controllable rule-based translation system spanning 50 English dialects and 189 unique linguistic features. Multi-VALUE maps SAE to synthetic forms of each dialect. First, we use this system to stress tests question answering, machine translation, and semantic parsing. Stress tests reveal significant performance disparities for leading models on non-standard dialects. Second, we use this system as a data augmentation technique to improve the dialect robustness of existing systems. Finally, we partner with native speakers of Chicano and Indian English to release new gold-standard variants of the popular CoQA task. To execute the transformation code, run model checkpoints, and download both synthetic and gold-standard dialectal benchmark datasets, see \url{http://value-nlp.org}.",
}
| Dialect differences caused by regional, social, and economic factors cause performance discrepancies for many groups of language technology users. Inclusive and equitable language technology must critically be dialect invariant, meaning that performance remains constant over dialectal shifts. Current systems often fall short of this ideal since they are designed and tested on a single dialect: Standard American English (SAE). We introduce a suite of resources for evaluating and achieving English dialect invariance. The resource is called Multi-VALUE, a controllable rule-based translation system spanning 50 English dialects and 189 unique linguistic features. Multi-VALUE maps SAE to synthetic forms of each dialect. First, we use this system to stress tests question answering, machine translation, and semantic parsing. Stress tests reveal significant performance disparities for leading models on non-standard dialects. Second, we use this system as a data augmentation technique to improve the dialect robustness of existing systems. Finally, we partner with native speakers of Chicano and Indian English to release new gold-standard variants of the popular CoQA task. To execute the transformation code, run model checkpoints, and download both synthetic and gold-standard dialectal benchmark datasets, see \url{http://value-nlp.org}. | [
"Ziems, Caleb",
"Held, William",
"Yang, Jingfeng",
"Dhamala, Jwala",
"Gupta, Rahul",
"Yang, Diyi"
] | Multi-VALUE: A Framework for Cross-Dialectal English NLP | acl-long.44 | Poster | 2212.08011 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.45.bib | https://aclanthology.org/2023.acl-long.45/ | @inproceedings{zhang-etal-2023-self,
title = "Self-Edit: Fault-Aware Code Editor for Code Generation",
author = "Zhang, Kechi and
Li, Zhuo and
Li, Jia and
Li, Ge and
Jin, Zhi",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.45",
doi = "10.18653/v1/2023.acl-long.45",
pages = "769--787",
abstract = "Large language models (LLMs) have demonstrated an impressive ability to generate codes on competitive programming tasks. However, with limited sample numbers, LLMs still suffer from poor accuracy. Inspired by the process of human programming, we propose a generate-and-edit approach named Self-Edit that utilizes execution results of the generated code from LLMs to improve the code quality on the competitive programming task. We execute the generated code on the example test case provided in the question and wrap execution results into a supplementary comment. Utilizing this comment as guidance, our fault-aware code editor is employed to correct errors in the generated code. We perform extensive evaluations across two competitive programming datasets with nine different LLMs. Compared to directly generating from LLMs, our approach can improve the average of pass@1 by 89{\%} on APPS-dev, 31{\%} on APPS-test, and 48{\%} on HumanEval over nine popular code generation LLMs with parameter sizes ranging from 110M to 175B. Compared to other post-processing methods, our method demonstrates superior accuracy and efficiency.",
}
| Large language models (LLMs) have demonstrated an impressive ability to generate codes on competitive programming tasks. However, with limited sample numbers, LLMs still suffer from poor accuracy. Inspired by the process of human programming, we propose a generate-and-edit approach named Self-Edit that utilizes execution results of the generated code from LLMs to improve the code quality on the competitive programming task. We execute the generated code on the example test case provided in the question and wrap execution results into a supplementary comment. Utilizing this comment as guidance, our fault-aware code editor is employed to correct errors in the generated code. We perform extensive evaluations across two competitive programming datasets with nine different LLMs. Compared to directly generating from LLMs, our approach can improve the average of pass@1 by 89{\%} on APPS-dev, 31{\%} on APPS-test, and 48{\%} on HumanEval over nine popular code generation LLMs with parameter sizes ranging from 110M to 175B. Compared to other post-processing methods, our method demonstrates superior accuracy and efficiency. | [
"Zhang, Kechi",
"Li, Zhuo",
"Li, Jia",
"Li, Ge",
"Jin, Zhi"
] | Self-Edit: Fault-Aware Code Editor for Code Generation | acl-long.45 | Poster | 2305.04087 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.46.bib | https://aclanthology.org/2023.acl-long.46/ | @inproceedings{don-yehiya-etal-2023-cold,
title = "{C}ol{D} Fusion: Collaborative Descent for Distributed Multitask Finetuning",
author = "Don-Yehiya, Shachar and
Venezian, Elad and
Raffel, Colin and
Slonim, Noam and
Choshen, Leshem",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.46",
doi = "10.18653/v1/2023.acl-long.46",
pages = "788--806",
abstract = "Pretraining has been shown to scale well with compute, data size and data diversity. Multitask learning trains on a mixture of supervised datasets and produces improved performance compared to self-supervised pretraining. Until now, massively multitask learning required simultaneous access to all datasets in the mixture and heavy compute resources that are only available to well-resourced teams. In this paper, we propose ColD Fusion, a method that provides the benefits of multitask learning but leverages distributed computation and requires limited communication and no sharing of data. Consequentially, ColD Fusion can create a synergistic loop, where finetuned models can be recycled to continually improve the pretrained model they are based on. We show that ColD Fusion yields comparable benefits to multitask training by producing a model that (a) attains strong performance on all of the datasets it was multitask trained on and (b) is a better starting point for finetuning on unseen datasets. We find ColD Fusion outperforms RoBERTa and even previous multitask models. Specifically, when training and testing on 35 diverse datasets, ColD Fusion-based model outperforms RoBERTa by 2.19 points on average without any changes to the architecture.",
}
| Pretraining has been shown to scale well with compute, data size and data diversity. Multitask learning trains on a mixture of supervised datasets and produces improved performance compared to self-supervised pretraining. Until now, massively multitask learning required simultaneous access to all datasets in the mixture and heavy compute resources that are only available to well-resourced teams. In this paper, we propose ColD Fusion, a method that provides the benefits of multitask learning but leverages distributed computation and requires limited communication and no sharing of data. Consequentially, ColD Fusion can create a synergistic loop, where finetuned models can be recycled to continually improve the pretrained model they are based on. We show that ColD Fusion yields comparable benefits to multitask training by producing a model that (a) attains strong performance on all of the datasets it was multitask trained on and (b) is a better starting point for finetuning on unseen datasets. We find ColD Fusion outperforms RoBERTa and even previous multitask models. Specifically, when training and testing on 35 diverse datasets, ColD Fusion-based model outperforms RoBERTa by 2.19 points on average without any changes to the architecture. | [
"Don-Yehiya, Shachar",
"Venezian, Elad",
"Raffel, Colin",
"Slonim, Noam",
"Choshen, Leshem"
] | ColD Fusion: Collaborative Descent for Distributed Multitask Finetuning | acl-long.46 | Poster | 2212.01378 | [
""
] | https://huggingface.co./papers/2212.01378 | 3 | 1 | 0 | 6 | 1 | [
"ibm/ColD-Fusion",
"ibm/ColD-Fusion-itr10-seed2",
"ibm/ColD-Fusion-itr13-seed0",
"ibm/ColD-Fusion-itr2-seed2",
"ibm/ColD-Fusion-itr21-seed2",
"ibm/ColD-Fusion-itr17-seed2",
"ibm/ColD-Fusion-itr18-seed2",
"ibm/ColD-Fusion-bert-base-uncased-itr14-seed0",
"ibm/ColD-Fusion-bert-base-uncased-itr15-seed0",
"ibm/ColD-Fusion-itr17-seed4",
"ibm/ColD-Fusion-itr14-seed3",
"ibm/ColD-Fusion-itr16-seed3",
"ibm/ColD-Fusion-itr9-seed3",
"ibm/ColD-Fusion-itr22-seed0",
"ibm/ColD-Fusion-bert-base-uncased-itr7-seed0",
"ibm/ColD-Fusion-itr14-seed1",
"ibm/ColD-Fusion-itr14-seed0",
"ibm/ColD-Fusion-itr1-seed0",
"ibm/ColD-Fusion-bert-base-uncased-itr3-seed0",
"ibm/ColD-Fusion-itr0-seed0",
"ibm/ColD-Fusion-bert-base-uncased-itr29-seed0",
"ibm/ColD-Fusion-itr11-seed2",
"ibm/ColD-Fusion-bert-base-uncased-itr20-seed0",
"ibm/ColD-Fusion-bert-base-uncased-itr9-seed0",
"ibm/ColD-Fusion-itr18-seed3",
"ibm/ColD-Fusion-itr18-seed0",
"ibm/ColD-Fusion-itr13-seed2",
"ibm/ColD-Fusion-itr9-seed1",
"ibm/ColD-Fusion-itr17-seed0",
"ibm/ColD-Fusion-bert-base-uncased-itr4-seed0",
"ibm/ColD-Fusion-itr18-seed4",
"ibm/ColD-Fusion-itr14-seed4",
"ibm/ColD-Fusion-itr27-seed3",
"ibm/ColD-Fusion-bert-base-uncased-itr10-seed0",
"ibm/ColD-Fusion-itr16-seed0",
"ibm/ColD-Fusion-itr12-seed0",
"ibm/ColD-Fusion-itr12-seed3",
"ibm/ColD-Fusion-bert-base-uncased-itr6-seed0",
"ibm/ColD-Fusion-itr22-seed2",
"ibm/ColD-Fusion-itr15-seed0",
"ibm/ColD-Fusion-itr26-seed4",
"ibm/ColD-Fusion-bert-base-uncased-itr17-seed0",
"ibm/ColD-Fusion-itr25-seed0",
"ibm/ColD-Fusion-itr12-seed2",
"ibm/ColD-Fusion-itr22-seed4",
"ibm/ColD-Fusion-bert-base-uncased-itr18-seed0",
"ibm/ColD-Fusion-itr10-seed1",
"ibm/ColD-Fusion-bert-base-uncased-itr8-seed0",
"ibm/ColD-Fusion-itr12-seed1",
"ibm/ColD-Fusion-bert-base-uncased-itr19-seed0",
"ibm/ColD-Fusion-itr16-seed1",
"ibm/ColD-Fusion-bert-base-uncased-itr22-seed0",
"ibm/ColD-Fusion-itr18-seed1",
"ibm/ColD-Fusion-bert-base-uncased-itr25-seed0",
"ibm/ColD-Fusion-bert-base-uncased-itr23-seed0",
"ibm/ColD-Fusion-itr9-seed4",
"ibm/ColD-Fusion-itr15-seed4",
"ibm/ColD-Fusion-itr11-seed3",
"ibm/ColD-Fusion-itr9-seed2",
"ibm/ColD-Fusion-itr17-seed3",
"ibm/ColD-Fusion-bert-base-uncased-itr26-seed0",
"ibm/ColD-Fusion-itr13-seed1",
"ibm/ColD-Fusion-itr9-seed0",
"ibm/ColD-Fusion-itr20-seed3",
"ibm/ColD-Fusion-bert-base-uncased-itr2-seed0",
"ibm/ColD-Fusion-itr21-seed3",
"ibm/ColD-Fusion-itr14-seed2",
"ibm/ColD-Fusion-itr1-seed4",
"ibm/ColD-Fusion-bert-base-uncased-itr16-seed0",
"ibm/ColD-Fusion-itr25-seed3",
"ibm/ColD-Fusion-itr15-seed2",
"ibm/ColD-Fusion-itr11-seed1",
"ibm/ColD-Fusion-bert-base-uncased-itr12-seed0",
"ibm/ColD-Fusion-itr13-seed4",
"ibm/ColD-Fusion-itr1-seed2",
"ibm/ColD-Fusion-itr12-seed4",
"ibm/ColD-Fusion-bert-base-uncased-itr13-seed0",
"ibm/ColD-Fusion-itr11-seed4",
"ibm/ColD-Fusion-itr10-seed3",
"ibm/ColD-Fusion-itr15-seed3",
"ibm/ColD-Fusion-bert-base-uncased-itr27-seed0",
"ibm/ColD-Fusion-itr1-seed3",
"ibm/ColD-Fusion-itr15-seed1",
"ibm/ColD-Fusion-itr23-seed3",
"ibm/ColD-Fusion-bert-base-uncased-itr11-seed0",
"ibm/ColD-Fusion-itr22-seed1",
"ibm/ColD-Fusion-itr16-seed4",
"ibm/ColD-Fusion-itr28-seed2",
"ibm/ColD-Fusion-bert-base-uncased-itr28-seed0",
"ibm/ColD-Fusion-itr7-seed4",
"ibm/ColD-Fusion-itr11-seed0",
"ibm/ColD-Fusion-itr2-seed0",
"ibm/ColD-Fusion-bert-base-uncased-itr0-seed0",
"ibm/ColD-Fusion-itr16-seed2",
"ibm/ColD-Fusion-bert-base-uncased-itr1-seed0",
"ibm/ColD-Fusion-bert-base-uncased-itr24-seed0",
"ibm/ColD-Fusion-bert-base-uncased-itr5-seed0",
"ibm/ColD-Fusion-bert-base-uncased-itr21-seed0",
"ibm/ColD-Fusion-itr1-seed1",
"ibm/ColD-Fusion-itr10-seed0"
] | [] | [] |
https://aclanthology.org/2023.acl-long.47.bib | https://aclanthology.org/2023.acl-long.47/ | @inproceedings{zhan-etal-2023-test,
title = "Test-time Adaptation for Machine Translation Evaluation by Uncertainty Minimization",
author = "Zhan, Runzhe and
Liu, Xuebo and
Wong, Derek F. and
Zhang, Cuilian and
Chao, Lidia S. and
Zhang, Min",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.47",
doi = "10.18653/v1/2023.acl-long.47",
pages = "807--820",
abstract = "The neural metrics recently received considerable attention from the research community in the automatic evaluation of machine translation. Unlike text-based metrics that have interpretable and consistent evaluation mechanisms for various data sources, the reliability of neural metrics in assessing out-of-distribution data remains a concern due to the disparity between training data and real-world data. This paper aims to address the inference bias of neural metrics through uncertainty minimization during test time, without requiring additional data. Our proposed method comprises three steps: uncertainty estimation, test-time adaptation, and inference. Specifically, the model employs the prediction uncertainty of the current data as a signal to update a small fraction of parameters during test time and subsequently refine the prediction through optimization. To validate our approach, we apply the proposed method to three representative models and conduct experiments on the WMT21 benchmarks. The results obtained from both in-domain and out-of-distribution evaluations consistently demonstrate improvements in correlation performance across different models. Furthermore, we provide evidence that the proposed method effectively reduces model uncertainty. The code is publicly available at \url{https://github.com/NLP2CT/TaU}.",
}
| The neural metrics recently received considerable attention from the research community in the automatic evaluation of machine translation. Unlike text-based metrics that have interpretable and consistent evaluation mechanisms for various data sources, the reliability of neural metrics in assessing out-of-distribution data remains a concern due to the disparity between training data and real-world data. This paper aims to address the inference bias of neural metrics through uncertainty minimization during test time, without requiring additional data. Our proposed method comprises three steps: uncertainty estimation, test-time adaptation, and inference. Specifically, the model employs the prediction uncertainty of the current data as a signal to update a small fraction of parameters during test time and subsequently refine the prediction through optimization. To validate our approach, we apply the proposed method to three representative models and conduct experiments on the WMT21 benchmarks. The results obtained from both in-domain and out-of-distribution evaluations consistently demonstrate improvements in correlation performance across different models. Furthermore, we provide evidence that the proposed method effectively reduces model uncertainty. The code is publicly available at \url{https://github.com/NLP2CT/TaU}. | [
"Zhan, Runzhe",
"Liu, Xuebo",
"Wong, Derek F.",
"Zhang, Cuilian",
"Chao, Lidia S.",
"Zhang, Min"
] | Test-time Adaptation for Machine Translation Evaluation by Uncertainty Minimization | acl-long.47 | Oral | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.48.bib | https://aclanthology.org/2023.acl-long.48/ | @inproceedings{chang-etal-2023-multi,
title = "Multi-{CLS} {BERT}: An Efficient Alternative to Traditional Ensembling",
author = "Chang, Haw-Shiuan and
Sun, Ruei-Yao and
Ricci, Kathryn and
McCallum, Andrew",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.48",
doi = "10.18653/v1/2023.acl-long.48",
pages = "821--854",
abstract = "Ensembling BERT models often significantly improves accuracy, but at the cost of significantly more computation and memory footprint. In this work, we propose Multi-CLS BERT, a novel ensembling method for CLS-based prediction tasks that is almost as efficient as a single BERT model. Multi-CLS BERT uses multiple CLS tokens with a parameterization and objective that encourages their diversity. Thus instead of fine-tuning each BERT model in an ensemble (and running them all at test time), we need only fine-tune our single Multi-CLS BERT model (and run the one model at test time, ensembling just the multiple final CLS embeddings). To test its effectiveness, we build Multi-CLS BERT on top of a state-of-the-art pretraining method for BERT (Aroca-Ouellette and Rudzicz, 2020). In experiments on GLUE and SuperGLUE we show that our Multi-CLS BERT reliably improves both overall accuracy and confidence estimation. When only 100 training samples are available in GLUE, the Multi-CLS BERT{\_}Base model can even outperform the corresponding BERT{\_}Large model. We analyze the behavior of our Multi-CLS BERT, showing that it has many of the same characteristics and behavior as a typical BERT 5-way ensemble, but with nearly 4-times less computation and memory.",
}
| Ensembling BERT models often significantly improves accuracy, but at the cost of significantly more computation and memory footprint. In this work, we propose Multi-CLS BERT, a novel ensembling method for CLS-based prediction tasks that is almost as efficient as a single BERT model. Multi-CLS BERT uses multiple CLS tokens with a parameterization and objective that encourages their diversity. Thus instead of fine-tuning each BERT model in an ensemble (and running them all at test time), we need only fine-tune our single Multi-CLS BERT model (and run the one model at test time, ensembling just the multiple final CLS embeddings). To test its effectiveness, we build Multi-CLS BERT on top of a state-of-the-art pretraining method for BERT (Aroca-Ouellette and Rudzicz, 2020). In experiments on GLUE and SuperGLUE we show that our Multi-CLS BERT reliably improves both overall accuracy and confidence estimation. When only 100 training samples are available in GLUE, the Multi-CLS BERT{\_}Base model can even outperform the corresponding BERT{\_}Large model. We analyze the behavior of our Multi-CLS BERT, showing that it has many of the same characteristics and behavior as a typical BERT 5-way ensemble, but with nearly 4-times less computation and memory. | [
"Chang, Haw-Shiuan",
"Sun, Ruei-Yao",
"Ricci, Kathryn",
"McCallum, Andrew"
] | Multi-CLS BERT: An Efficient Alternative to Traditional Ensembling | acl-long.48 | Poster | 2210.05043 | [
"https://github.com/iesl/multicls"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.49.bib | https://aclanthology.org/2023.acl-long.49/ | @inproceedings{ai-fang-2023-fly,
title = "On-the-fly Cross-lingual Masking for Multilingual Pre-training",
author = "Ai, Xi and
Fang, Bin",
editor = "Rogers, Anna and
Boyd-Graber, Jordan and
Okazaki, Naoaki",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.49",
doi = "10.18653/v1/2023.acl-long.49",
pages = "855--876",
abstract = "In multilingual pre-training with the objective of MLM (masked language modeling) on multiple monolingual corpora, multilingual models only learn cross-linguality implicitly from isomorphic spaces formed by overlapping different language spaces due to the lack of explicit cross-lingual forward pass. In this work, we present CLPM (Cross-lingual Prototype Masking), a dynamic and token-wise masking scheme, for multilingual pre-training, using a special token $[\mathcal{C}]_{x}$ to replace a random token $x$ in the input sentence. $[\mathcal{C}]_{x}$ is a cross-lingual prototype for $x$ and then forms an explicit cross-lingual forward pass. We instantiate CLPM for the multilingual pre-training phase of UNMT (unsupervised neural machine translation), and experiments show that CLPM can consistently improve the performance of UNMT models on $\{De, Ro, Ne \} \leftrightarrow En$. Beyond UNMT or bilingual tasks, we show that CLPM can consistently improve the performance of multilingual models on cross-lingual classification.",
}
| In multilingual pre-training with the objective of MLM (masked language modeling) on multiple monolingual corpora, multilingual models only learn cross-linguality implicitly from isomorphic spaces formed by overlapping different language spaces due to the lack of explicit cross-lingual forward pass. In this work, we present CLPM (Cross-lingual Prototype Masking), a dynamic and token-wise masking scheme, for multilingual pre-training, using a special token $[\mathcal{C}]_{x}$ to replace a random token $x$ in the input sentence. $[\mathcal{C}]_{x}$ is a cross-lingual prototype for $x$ and then forms an explicit cross-lingual forward pass. We instantiate CLPM for the multilingual pre-training phase of UNMT (unsupervised neural machine translation), and experiments show that CLPM can consistently improve the performance of UNMT models on $\{De, Ro, Ne \} \leftrightarrow En$. Beyond UNMT or bilingual tasks, we show that CLPM can consistently improve the performance of multilingual models on cross-lingual classification. | [
"Ai, Xi",
"Fang, Bin"
] | On-the-fly Cross-lingual Masking for Multilingual Pre-training | acl-long.49 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
End of preview. Expand
in Dataset Viewer.
README.md exists but content is empty.
- Downloads last month
- 108