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.300.bib | https://aclanthology.org/2023.acl-long.300/ | @inproceedings{onoe-etal-2023-lms,
title = "Can {LM}s Learn New Entities from Descriptions? Challenges in Propagating Injected Knowledge",
author = "Onoe, Yasumasa and
Zhang, Michael and
Padmanabhan, Shankar and
Durrett, Greg and
Choi, Eunsol",
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.300",
doi = "10.18653/v1/2023.acl-long.300",
pages = "5469--5485",
abstract = "Pre-trained language models (LMs) are used for knowledge intensive tasks like question answering, but their knowledge gets continuously outdated as the world changes. Prior work has studied targeted updates to LMs, injecting individual facts and evaluating whether the model learns these facts while not changing predictions on other contexts. We take a step forward and study LMs{'} abilities to make inferences based on injected facts (or propagate those facts): for example, after learning that something is a TV show, does an LM predict that you can watch it? We study this with two cloze-style tasks: an existing dataset of real-world sentences about novel entities (ECBD) as well as a new controlled benchmark with manually designed templates requiring varying levels of inference about injected knowledge. Surprisingly, we find that existing methods for updating knowledge (gradient-based fine-tuning and modifications of this approach) show little propagation of injected knowledge. These methods improve performance on cloze instances only when there is lexical overlap between injected facts and target inferences. Yet, prepending entity definitions in an LM{'}s context improves performance across all settings, suggesting that there is substantial headroom for parameter-updating approaches for knowledge injection.",
}
| Pre-trained language models (LMs) are used for knowledge intensive tasks like question answering, but their knowledge gets continuously outdated as the world changes. Prior work has studied targeted updates to LMs, injecting individual facts and evaluating whether the model learns these facts while not changing predictions on other contexts. We take a step forward and study LMs{'} abilities to make inferences based on injected facts (or propagate those facts): for example, after learning that something is a TV show, does an LM predict that you can watch it? We study this with two cloze-style tasks: an existing dataset of real-world sentences about novel entities (ECBD) as well as a new controlled benchmark with manually designed templates requiring varying levels of inference about injected knowledge. Surprisingly, we find that existing methods for updating knowledge (gradient-based fine-tuning and modifications of this approach) show little propagation of injected knowledge. These methods improve performance on cloze instances only when there is lexical overlap between injected facts and target inferences. Yet, prepending entity definitions in an LM{'}s context improves performance across all settings, suggesting that there is substantial headroom for parameter-updating approaches for knowledge injection. | [
"Onoe, Yasumasa",
"Zhang, Michael",
"Padmanabhan, Shankar",
"Durrett, Greg",
"Choi, Eunsol"
] | Can LMs Learn New Entities from Descriptions? Challenges in Propagating Injected Knowledge | acl-long.300 | Poster | 2305.01651 | [
"https://github.com/yasumasaonoe/entity_knowledge_propagation"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.301.bib | https://aclanthology.org/2023.acl-long.301/ | @inproceedings{ferrando-etal-2023-explaining,
title = "Explaining How Transformers Use Context to Build Predictions",
author = "Ferrando, Javier and
G{\'a}llego, Gerard I. and
Tsiamas, Ioannis 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.301",
doi = "10.18653/v1/2023.acl-long.301",
pages = "5486--5513",
abstract = "Language Generation Models produce words based on the previous context. Although existing methods offer input attributions as explanations for a model{'}s prediction, it is still unclear how prior words affect the model{'}s decision throughout the layers. In this work, we leverage recent advances in explainability of the Transformer and present a procedure to analyze models for language generation. Using contrastive examples, we compare the alignment of our explanations with evidence of the linguistic phenomena, and show that our method consistently aligns better than gradient-based and perturbation-based baselines. Then, we investigate the role of MLPs inside the Transformer and show that they learn features that help the model predict words that are grammatically acceptable. Lastly, we apply our method to Neural Machine Translation models, and demonstrate that they generate human-like source-target alignments for building predictions.",
}
| Language Generation Models produce words based on the previous context. Although existing methods offer input attributions as explanations for a model{'}s prediction, it is still unclear how prior words affect the model{'}s decision throughout the layers. In this work, we leverage recent advances in explainability of the Transformer and present a procedure to analyze models for language generation. Using contrastive examples, we compare the alignment of our explanations with evidence of the linguistic phenomena, and show that our method consistently aligns better than gradient-based and perturbation-based baselines. Then, we investigate the role of MLPs inside the Transformer and show that they learn features that help the model predict words that are grammatically acceptable. Lastly, we apply our method to Neural Machine Translation models, and demonstrate that they generate human-like source-target alignments for building predictions. | [
"Ferr",
"o, Javier",
"G{\\'a}llego, Gerard I.",
"Tsiamas, Ioannis",
"Costa-juss{\\`a}, Marta R."
] | Explaining How Transformers Use Context to Build Predictions | acl-long.301 | Poster | 2305.12535 | [
"https://github.com/mt-upc/logit-explanations"
] | https://huggingface.co/papers/2305.12535 | 1 | 0 | 0 | 4 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.302.bib | https://aclanthology.org/2023.acl-long.302/ | @inproceedings{chen-etal-2023-disco,
title = "{DISCO}: Distilling Counterfactuals with Large Language Models",
author = "Chen, Zeming and
Gao, Qiyue and
Bosselut, Antoine and
Sabharwal, Ashish and
Richardson, Kyle",
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.302",
doi = "10.18653/v1/2023.acl-long.302",
pages = "5514--5528",
abstract = "Models trained with counterfactually augmented data learn representations of the causal structure of tasks, enabling robust generalization. However, high-quality counterfactual data is scarce for most tasks and not easily generated at scale. When crowdsourced, such data is typically limited in scale and diversity; when generated using supervised methods, it is computationally expensive to extend to new counterfactual dimensions. In this work, we introduce DISCO (DIStilled COunterfactual Data), a new method for automatically generating high-quality counterfactual data at scale. DISCO engineers prompts to generate phrasal perturbations with a large general language model. Then, a task-specific teacher model filters these generations to distill high-quality counterfactual data. While task-agnostic, we apply our pipeline to the task of natural language inference (NLI) and find that on challenging evaluations such as the NLI stress test, comparatively smaller student models trained with DISCO generated counterfactuals are more robust (6{\%} absolute) and generalize better across distributions (2{\%}) compared to models trained without data augmentation. Furthermore, DISCO augmented models are 10{\%} more consistent between counterfactual pairs on three evaluation sets, demonstrating that DISCO augmentation enables models to more reliably learn causal representations. Our repository are available at: \url{https://github.com/eric11eca/disco}",
}
| Models trained with counterfactually augmented data learn representations of the causal structure of tasks, enabling robust generalization. However, high-quality counterfactual data is scarce for most tasks and not easily generated at scale. When crowdsourced, such data is typically limited in scale and diversity; when generated using supervised methods, it is computationally expensive to extend to new counterfactual dimensions. In this work, we introduce DISCO (DIStilled COunterfactual Data), a new method for automatically generating high-quality counterfactual data at scale. DISCO engineers prompts to generate phrasal perturbations with a large general language model. Then, a task-specific teacher model filters these generations to distill high-quality counterfactual data. While task-agnostic, we apply our pipeline to the task of natural language inference (NLI) and find that on challenging evaluations such as the NLI stress test, comparatively smaller student models trained with DISCO generated counterfactuals are more robust (6{\%} absolute) and generalize better across distributions (2{\%}) compared to models trained without data augmentation. Furthermore, DISCO augmented models are 10{\%} more consistent between counterfactual pairs on three evaluation sets, demonstrating that DISCO augmentation enables models to more reliably learn causal representations. Our repository are available at: \url{https://github.com/eric11eca/disco} | [
"Chen, Zeming",
"Gao, Qiyue",
"Bosselut, Antoine",
"Sabharwal, Ashish",
"Richardson, Kyle"
] | DISCO: Distilling Counterfactuals with Large Language Models | acl-long.302 | Poster | 2212.10534 | [
"https://github.com/eric11eca/disco"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.303.bib | https://aclanthology.org/2023.acl-long.303/ | @inproceedings{zhou-etal-2023-non,
title = "Non-Sequential Graph Script Induction via Multimedia Grounding",
author = "Zhou, Yu and
Li, Sha and
Li, Manling and
Lin, Xudong and
Chang, Shih-Fu and
Bansal, Mohit and
Ji, Heng",
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.303",
doi = "10.18653/v1/2023.acl-long.303",
pages = "5529--5545",
abstract = "Online resources such as WikiHow compile a wide range of scripts for performing everyday tasks, which can assist models in learning to reason about procedures. However, the scripts are always presented in a linear manner, which does not reflect the flexibility displayed by people executing tasks in real life. For example, in the CrossTask Dataset, 64.5{\%} of consecutive step pairs are also observed in the reverse order, suggesting their ordering is not fixed. In addition, each step has an average of 2.56 frequent next steps, demonstrating {``}branching{''}. In this paper, we propose the new challenging task of non-sequential graph script induction, aiming to capture optional and interchangeable steps in procedural planning. To automate the induction of such graph scripts for given tasks, we propose to take advantage of loosely aligned videos of people performing the tasks. In particular, we design a multimodal framework to ground procedural videos to WikiHow textual steps and thus transform each video into an observed step path on the latent ground truth graph script. This key transformation enables us to train a script knowledge model capable of both generating explicit graph scripts for learnt tasks and predicting future steps given a partial step sequence. Our best model outperforms the strongest pure text/vision baselines by 17.52{\%} absolute gains on F1@3 for next step prediction and 13.8{\%} absolute gains on Acc@1 for partial sequence completion. Human evaluation shows our model outperforming the WikiHow linear baseline by 48.76{\%} absolute gains in capturing sequential and non-sequential step relationships.",
}
| Online resources such as WikiHow compile a wide range of scripts for performing everyday tasks, which can assist models in learning to reason about procedures. However, the scripts are always presented in a linear manner, which does not reflect the flexibility displayed by people executing tasks in real life. For example, in the CrossTask Dataset, 64.5{\%} of consecutive step pairs are also observed in the reverse order, suggesting their ordering is not fixed. In addition, each step has an average of 2.56 frequent next steps, demonstrating {``}branching{''}. In this paper, we propose the new challenging task of non-sequential graph script induction, aiming to capture optional and interchangeable steps in procedural planning. To automate the induction of such graph scripts for given tasks, we propose to take advantage of loosely aligned videos of people performing the tasks. In particular, we design a multimodal framework to ground procedural videos to WikiHow textual steps and thus transform each video into an observed step path on the latent ground truth graph script. This key transformation enables us to train a script knowledge model capable of both generating explicit graph scripts for learnt tasks and predicting future steps given a partial step sequence. Our best model outperforms the strongest pure text/vision baselines by 17.52{\%} absolute gains on F1@3 for next step prediction and 13.8{\%} absolute gains on Acc@1 for partial sequence completion. Human evaluation shows our model outperforming the WikiHow linear baseline by 48.76{\%} absolute gains in capturing sequential and non-sequential step relationships. | [
"Zhou, Yu",
"Li, Sha",
"Li, Manling",
"Lin, Xudong",
"Chang, Shih-Fu",
"Bansal, Mohit",
"Ji, Heng"
] | Non-Sequential Graph Script Induction via Multimedia Grounding | acl-long.303 | Poster | 2305.17542 | [
"https://github.com/bryanzhou008/multimodal-graph-script-learning"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.304.bib | https://aclanthology.org/2023.acl-long.304/ | @inproceedings{wang-etal-2023-scott,
title = "{SCOTT}: Self-Consistent Chain-of-Thought Distillation",
author = "Wang, Peifeng and
Wang, Zhengyang and
Li, Zheng and
Gao, Yifan and
Yin, Bing and
Ren, Xiang",
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.304",
doi = "10.18653/v1/2023.acl-long.304",
pages = "5546--5558",
abstract = "Large language models (LMs) beyond a certain scale, demonstrate the emergent capability of generating free-text rationales for their predictions via chain-of-thought (CoT) prompting. While CoT can yield dramatically improved performance, such gains are only observed for sufficiently large LMs. Even more concerning, there is little guarantee that the generated rationales are consistent with LM{'}s predictions or faithfully justify the decisions. In this work, we propose SCOTT, a faithful knowledge distillation method to learn a small, self-consistent CoT model from a teacher model that is orders of magnitude larger. To form better supervision, we elicit rationales supporting the gold answers from a large LM (teacher) by contrastive decoding, which encourages the teacher to generate tokens that become more plausible only when the answer is considered. To ensure faithful distillation, we use the teacher-generated rationales to learn a student LM with a counterfactual reasoning objective, which prevents the student from ignoring the rationales to make inconsistent predictions. Experiments show that while yielding comparable performance, our method leads to a more faithful model than baselines. Further analysis shows that such a model respects the rationales more when making decisions; thus, we can improve its performance more by refining its rationales.",
}
| Large language models (LMs) beyond a certain scale, demonstrate the emergent capability of generating free-text rationales for their predictions via chain-of-thought (CoT) prompting. While CoT can yield dramatically improved performance, such gains are only observed for sufficiently large LMs. Even more concerning, there is little guarantee that the generated rationales are consistent with LM{'}s predictions or faithfully justify the decisions. In this work, we propose SCOTT, a faithful knowledge distillation method to learn a small, self-consistent CoT model from a teacher model that is orders of magnitude larger. To form better supervision, we elicit rationales supporting the gold answers from a large LM (teacher) by contrastive decoding, which encourages the teacher to generate tokens that become more plausible only when the answer is considered. To ensure faithful distillation, we use the teacher-generated rationales to learn a student LM with a counterfactual reasoning objective, which prevents the student from ignoring the rationales to make inconsistent predictions. Experiments show that while yielding comparable performance, our method leads to a more faithful model than baselines. Further analysis shows that such a model respects the rationales more when making decisions; thus, we can improve its performance more by refining its rationales. | [
"Wang, Peifeng",
"Wang, Zhengyang",
"Li, Zheng",
"Gao, Yifan",
"Yin, Bing",
"Ren, Xiang"
] | SCOTT: Self-Consistent Chain-of-Thought Distillation | acl-long.304 | Oral | 2305.01879 | [
"https://github.com/wangpf3/consistent-cot-distillation"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.305.bib | https://aclanthology.org/2023.acl-long.305/ | @inproceedings{kim-etal-2023-clinical,
title = "Clinical Note Owns its Hierarchy: Multi-Level Hypergraph Neural Networks for Patient-Level Representation Learning",
author = "Kim, Nayeon and
Piao, Yinhua and
Kim, Sun",
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.305",
doi = "10.18653/v1/2023.acl-long.305",
pages = "5559--5573",
abstract = "Leveraging knowledge from electronic health records (EHRs) to predict a patient{'}s condition is essential to the effective delivery of appropriate care. Clinical notes of patient EHRs contain valuable information from healthcare professionals, but have been underused due to their difficult contents and complex hierarchies. Recently, hypergraph-based methods have been proposed for document classifications. Directly adopting existing hypergraph methods on clinical notes cannot sufficiently utilize the hierarchy information of the patient, which can degrade clinical semantic information by (1) frequent neutral words and (2) hierarchies with imbalanced distribution. Thus, we propose a taxonomy-aware multi-level hypergraph neural network (TM-HGNN), where multi-level hypergraphs assemble useful neutral words with rare keywords via note and taxonomy level hyperedges to retain the clinical semantic information. The constructed patient hypergraphs are fed into hierarchical message passing layers for learning more balanced multi-level knowledge at the note and taxonomy levels. We validate the effectiveness of TM-HGNN by conducting extensive experiments with MIMIC-III dataset on benchmark in-hospital-mortality prediction.",
}
| Leveraging knowledge from electronic health records (EHRs) to predict a patient{'}s condition is essential to the effective delivery of appropriate care. Clinical notes of patient EHRs contain valuable information from healthcare professionals, but have been underused due to their difficult contents and complex hierarchies. Recently, hypergraph-based methods have been proposed for document classifications. Directly adopting existing hypergraph methods on clinical notes cannot sufficiently utilize the hierarchy information of the patient, which can degrade clinical semantic information by (1) frequent neutral words and (2) hierarchies with imbalanced distribution. Thus, we propose a taxonomy-aware multi-level hypergraph neural network (TM-HGNN), where multi-level hypergraphs assemble useful neutral words with rare keywords via note and taxonomy level hyperedges to retain the clinical semantic information. The constructed patient hypergraphs are fed into hierarchical message passing layers for learning more balanced multi-level knowledge at the note and taxonomy levels. We validate the effectiveness of TM-HGNN by conducting extensive experiments with MIMIC-III dataset on benchmark in-hospital-mortality prediction. | [
"Kim, Nayeon",
"Piao, Yinhua",
"Kim, Sun"
] | Clinical Note Owns its Hierarchy: Multi-Level Hypergraph Neural Networks for Patient-Level Representation Learning | acl-long.305 | Oral | 2305.09756 | [
"https://github.com/ny1031/tm-hgnn"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.306.bib | https://aclanthology.org/2023.acl-long.306/ | @inproceedings{pu-etal-2023-incorporating,
title = "Incorporating Distributions of Discourse Structure for Long Document Abstractive Summarization",
author = "Pu, Dongqi and
Wang, Yifan and
Demberg, Vera",
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.306",
doi = "10.18653/v1/2023.acl-long.306",
pages = "5574--5590",
abstract = "For text summarization, the role of discourse structure is pivotal in discerning the core content of a text. Regrettably, prior studies on incorporating Rhetorical Structure Theory (RST) into transformer-based summarization models only consider the nuclearity annotation, thereby overlooking the variety of discourse relation types. This paper introduces the {`}RSTformer{'}, a novel summarization model that comprehensively incorporates both the types and uncertainty of rhetorical relations. Our RST-attention mechanism, rooted in document-level rhetorical structure, is an extension of the recently devised Longformer framework. Through rigorous evaluation, the model proposed herein exhibits significant superiority over state-of-the-art models, as evidenced by its notable performance on several automatic metrics and human evaluation.",
}
| For text summarization, the role of discourse structure is pivotal in discerning the core content of a text. Regrettably, prior studies on incorporating Rhetorical Structure Theory (RST) into transformer-based summarization models only consider the nuclearity annotation, thereby overlooking the variety of discourse relation types. This paper introduces the {`}RSTformer{'}, a novel summarization model that comprehensively incorporates both the types and uncertainty of rhetorical relations. Our RST-attention mechanism, rooted in document-level rhetorical structure, is an extension of the recently devised Longformer framework. Through rigorous evaluation, the model proposed herein exhibits significant superiority over state-of-the-art models, as evidenced by its notable performance on several automatic metrics and human evaluation. | [
"Pu, Dongqi",
"Wang, Yifan",
"Demberg, Vera"
] | Incorporating Distributions of Discourse Structure for Long Document Abstractive Summarization | acl-long.306 | Poster | 2305.16784 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.307.bib | https://aclanthology.org/2023.acl-long.307/ | @inproceedings{kamalloo-etal-2023-evaluating,
title = "Evaluating Open-Domain Question Answering in the Era of Large Language Models",
author = "Kamalloo, Ehsan and
Dziri, Nouha and
Clarke, Charles and
Rafiei, Davood",
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.307",
doi = "10.18653/v1/2023.acl-long.307",
pages = "5591--5606",
abstract = "Lexical matching remains the de facto evaluation method for open-domain question answering (QA). Unfortunately, lexical matching fails completely when a plausible candidate answer does not appear in the list of gold answers, which is increasingly the case as we shift from extractive to generative models. The recent success of large language models (LLMs) for QA aggravates lexical matching failures since candidate answers become longer, thereby making matching with the gold answers even more challenging. Without accurate evaluation, the true progress in open-domain QA remains unknown. In this paper, we conduct a thorough analysis of various open-domain QA models, including LLMs, by manually evaluating their answers on a subset of NQ-open, a popular benchmark. Our assessments reveal that while the true performance of all models is significantly underestimated, the performance of the InstructGPT (zero-shot) LLM increases by nearly +60{\%}, making it on par with existing top models, and the InstructGPT (few-shot) model actually achieves a new state-of-the-art on NQ-open. We also find that more than 50{\%} of lexical matching failures are attributed to semantically equivalent answers. We further demonstrate that regex matching ranks QA models consistent with human judgments, although still suffering from unnecessary strictness. Finally, we demonstrate that automated evaluation models are a reasonable surrogate for lexical matching in some circumstances, but not for long-form answers generated by LLMs. The automated models struggle in detecting hallucinations in LLM answers and are thus unable to evaluate LLMs. At this time, there appears to be no substitute for human evaluation.",
}
| Lexical matching remains the de facto evaluation method for open-domain question answering (QA). Unfortunately, lexical matching fails completely when a plausible candidate answer does not appear in the list of gold answers, which is increasingly the case as we shift from extractive to generative models. The recent success of large language models (LLMs) for QA aggravates lexical matching failures since candidate answers become longer, thereby making matching with the gold answers even more challenging. Without accurate evaluation, the true progress in open-domain QA remains unknown. In this paper, we conduct a thorough analysis of various open-domain QA models, including LLMs, by manually evaluating their answers on a subset of NQ-open, a popular benchmark. Our assessments reveal that while the true performance of all models is significantly underestimated, the performance of the InstructGPT (zero-shot) LLM increases by nearly +60{\%}, making it on par with existing top models, and the InstructGPT (few-shot) model actually achieves a new state-of-the-art on NQ-open. We also find that more than 50{\%} of lexical matching failures are attributed to semantically equivalent answers. We further demonstrate that regex matching ranks QA models consistent with human judgments, although still suffering from unnecessary strictness. Finally, we demonstrate that automated evaluation models are a reasonable surrogate for lexical matching in some circumstances, but not for long-form answers generated by LLMs. The automated models struggle in detecting hallucinations in LLM answers and are thus unable to evaluate LLMs. At this time, there appears to be no substitute for human evaluation. | [
"Kamalloo, Ehsan",
"Dziri, Nouha",
"Clarke, Charles",
"Rafiei, Davood"
] | Evaluating Open-Domain Question Answering in the Era of Large Language Models | acl-long.307 | Oral | 2305.06984 | [
"https://github.com/ehsk/openqa-eval"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.308.bib | https://aclanthology.org/2023.acl-long.308/ | @inproceedings{pitarch-etal-2023-clues,
title = "No clues good clues: out of context Lexical Relation Classification",
author = "Pitarch, Lucia and
Bernad, Jordi and
Dranca, Lacramioara and
Bobed Lisbona, Carlos and
Gracia, Jorge",
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.308",
doi = "10.18653/v1/2023.acl-long.308",
pages = "5607--5625",
abstract = "The accurate prediction of lexical relations between words is a challenging task in Natural Language Processing (NLP). The most recent advances in this direction come with the use of pre-trained language models (PTLMs). A PTLM typically needs {``}well-formed{''} verbalized text to interact with it, either to fine-tune it or to exploit it. However, there are indications that commonly used PTLMs already encode enough linguistic knowledge to allow the use of minimal (or none) textual context for some linguistically motivated tasks, thus notably reducing human effort, the need for data pre-processing, and favoring techniques that are language neutral since do not rely on syntactic structures. In this work, we explore this idea for the tasks of lexical relation classification (LRC) and graded Lexical Entailment (LE). After fine-tuning PTLMs for LRC with different verbalizations, our evaluation results show that very simple prompts are competitive for LRC and significantly outperform graded LE SoTA. In order to gain a better insight into this phenomenon, we perform a number of quantitative statistical analyses on the results, as well as a qualitative visual exploration based on embedding projections.",
}
| The accurate prediction of lexical relations between words is a challenging task in Natural Language Processing (NLP). The most recent advances in this direction come with the use of pre-trained language models (PTLMs). A PTLM typically needs {``}well-formed{''} verbalized text to interact with it, either to fine-tune it or to exploit it. However, there are indications that commonly used PTLMs already encode enough linguistic knowledge to allow the use of minimal (or none) textual context for some linguistically motivated tasks, thus notably reducing human effort, the need for data pre-processing, and favoring techniques that are language neutral since do not rely on syntactic structures. In this work, we explore this idea for the tasks of lexical relation classification (LRC) and graded Lexical Entailment (LE). After fine-tuning PTLMs for LRC with different verbalizations, our evaluation results show that very simple prompts are competitive for LRC and significantly outperform graded LE SoTA. In order to gain a better insight into this phenomenon, we perform a number of quantitative statistical analyses on the results, as well as a qualitative visual exploration based on embedding projections. | [
"Pitarch, Lucia",
"Bernad, Jordi",
"Dranca, Lacramioara",
"Bobed Lisbona, Carlos",
"Gracia, Jorge"
] | No clues good clues: out of context Lexical Relation Classification | acl-long.308 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.309.bib | https://aclanthology.org/2023.acl-long.309/ | @inproceedings{hu-etal-2023-wont,
title = "Won{'}t Get Fooled Again: Answering Questions with False Premises",
author = "Hu, Shengding and
Luo, Yifan and
Wang, Huadong and
Cheng, Xingyi and
Liu, Zhiyuan and
Sun, Maosong",
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.309",
doi = "10.18653/v1/2023.acl-long.309",
pages = "5626--5643",
abstract = "Pre-trained language models (PLMs) have shown unprecedented potential in various fields, especially as the backbones for question-answering (QA) systems. However, they tend to be easily deceived by tricky questions such as {``}How many eyes does the sun have?{''}. Such frailties of PLMs often allude to the lack of knowledge within them. In this paper, we find that the PLMs already possess the knowledge required to rebut such questions, and the key is how to activate the knowledge. To systematize this observation, we investigate the PLMs{'} responses to one kind of tricky questions, i.e., the false premises questions (FPQs). We annotate a FalseQA dataset containing 2365 human-written FPQs, with the corresponding explanations for the false premises and the revised true premise questions. Using FalseQA, we discover that PLMs are capable of discriminating FPQs by fine-tuning on moderate numbers (e.g., 256) of examples. PLMs also generate reasonable explanations for the false premise, which serve as rebuttals. Further replaying a few general questions during training allows PLMs to excel on FPQs and general questions simultaneously. Our work suggests that once the rebuttal ability is stimulated, knowledge inside the PLMs can be effectively utilized to handle FPQs, which incentivizes the research on PLM-based QA systems. The FalseQA dataset and code are available at \url{https://github.com/thunlp/FalseQA} .",
}
| Pre-trained language models (PLMs) have shown unprecedented potential in various fields, especially as the backbones for question-answering (QA) systems. However, they tend to be easily deceived by tricky questions such as {``}How many eyes does the sun have?{''}. Such frailties of PLMs often allude to the lack of knowledge within them. In this paper, we find that the PLMs already possess the knowledge required to rebut such questions, and the key is how to activate the knowledge. To systematize this observation, we investigate the PLMs{'} responses to one kind of tricky questions, i.e., the false premises questions (FPQs). We annotate a FalseQA dataset containing 2365 human-written FPQs, with the corresponding explanations for the false premises and the revised true premise questions. Using FalseQA, we discover that PLMs are capable of discriminating FPQs by fine-tuning on moderate numbers (e.g., 256) of examples. PLMs also generate reasonable explanations for the false premise, which serve as rebuttals. Further replaying a few general questions during training allows PLMs to excel on FPQs and general questions simultaneously. Our work suggests that once the rebuttal ability is stimulated, knowledge inside the PLMs can be effectively utilized to handle FPQs, which incentivizes the research on PLM-based QA systems. The FalseQA dataset and code are available at \url{https://github.com/thunlp/FalseQA} . | [
"Hu, Shengding",
"Luo, Yifan",
"Wang, Huadong",
"Cheng, Xingyi",
"Liu, Zhiyuan",
"Sun, Maosong"
] | Won't Get Fooled Again: Answering Questions with False Premises | acl-long.309 | Oral | 2307.02394 | [
"https://github.com/thunlp/falseqa"
] | https://huggingface.co/papers/2307.02394 | 1 | 0 | 0 | 6 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.310.bib | https://aclanthology.org/2023.acl-long.310/ | @inproceedings{tang-etal-2023-daam,
title = "What the {DAAM}: Interpreting Stable Diffusion Using Cross Attention",
author = "Tang, Raphael and
Liu, Linqing and
Pandey, Akshat and
Jiang, Zhiying and
Yang, Gefei and
Kumar, Karun and
Stenetorp, Pontus and
Lin, Jimmy and
Ture, Ferhan",
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.310",
doi = "10.18653/v1/2023.acl-long.310",
pages = "5644--5659",
abstract = "Diffusion models are a milestone in text-to-image generation, but they remain poorly understood, lacking interpretability analyses. In this paper, we perform a text-image attribution analysis on Stable Diffusion, a recently open-sourced model. To produce attribution maps, we upscale and aggregate cross-attention maps in the denoising module, naming our method DAAM. We validate it by testing its segmentation ability on nouns, as well as its generalized attribution quality on all parts of speech, rated by humans. On two generated datasets, we attain a competitive 58.8-64.8 mIoU on noun segmentation and fair to good mean opinion scores (3.4-4.2) on generalized attribution. Then, we apply DAAM to study the role of syntax in the pixel space across head{--}dependent heat map interaction patterns for ten common dependency relations. We show that, for some relations, the head map consistently subsumes the dependent, while the opposite is true for others. Finally, we study several semantic phenomena, focusing on feature entanglement; we find that the presence of cohyponyms worsens generation quality by 9{\%}, and descriptive adjectives attend too broadly. We are the first to interpret large diffusion models from a visuolinguistic perspective, which enables future research. Our code is at \url{https://github.com/castorini/daam}.",
}
| Diffusion models are a milestone in text-to-image generation, but they remain poorly understood, lacking interpretability analyses. In this paper, we perform a text-image attribution analysis on Stable Diffusion, a recently open-sourced model. To produce attribution maps, we upscale and aggregate cross-attention maps in the denoising module, naming our method DAAM. We validate it by testing its segmentation ability on nouns, as well as its generalized attribution quality on all parts of speech, rated by humans. On two generated datasets, we attain a competitive 58.8-64.8 mIoU on noun segmentation and fair to good mean opinion scores (3.4-4.2) on generalized attribution. Then, we apply DAAM to study the role of syntax in the pixel space across head{--}dependent heat map interaction patterns for ten common dependency relations. We show that, for some relations, the head map consistently subsumes the dependent, while the opposite is true for others. Finally, we study several semantic phenomena, focusing on feature entanglement; we find that the presence of cohyponyms worsens generation quality by 9{\%}, and descriptive adjectives attend too broadly. We are the first to interpret large diffusion models from a visuolinguistic perspective, which enables future research. Our code is at \url{https://github.com/castorini/daam}. | [
"Tang, Raphael",
"Liu, Linqing",
"P",
"ey, Akshat",
"Jiang, Zhiying",
"Yang, Gefei",
"Kumar, Karun",
"Stenetorp, Pontus",
"Lin, Jimmy",
"Ture, Ferhan"
] | What the DAAM: Interpreting Stable Diffusion Using Cross Attention | acl-long.310 | Poster | 2210.04885 | [
"https://github.com/castorini/daam"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.311.bib | https://aclanthology.org/2023.acl-long.311/ | @inproceedings{huang-etal-2023-zero,
title = "Zero-shot Faithful Factual Error Correction",
author = "Huang, Kung-Hsiang and
Chan, Hou Pong and
Ji, Heng",
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.311",
doi = "10.18653/v1/2023.acl-long.311",
pages = "5660--5676",
abstract = "Faithfully correcting factual errors is critical for maintaining the integrity of textual knowledge bases and preventing hallucinations in sequence-to-sequence models. Drawing on humans{'} ability to identify and correct factual errors, we present a zero-shot framework that formulates questions about input claims, looks for correct answers in the given evidence, and assesses the faithfulness of each correction based on its consistency with the evidence. Our zero-shot framework outperforms fully-supervised approaches, as demonstrated by experiments on the FEVER and SciFact datasets, where our outputs are shown to be more faithful. More importantly, the decomposability nature of our framework inherently provides interpretability. Additionally, to reveal the most suitable metrics for evaluating factual error corrections, we analyze the correlation between commonly used metrics with human judgments in terms of three different dimensions regarding intelligibility and faithfulness.",
}
| Faithfully correcting factual errors is critical for maintaining the integrity of textual knowledge bases and preventing hallucinations in sequence-to-sequence models. Drawing on humans{'} ability to identify and correct factual errors, we present a zero-shot framework that formulates questions about input claims, looks for correct answers in the given evidence, and assesses the faithfulness of each correction based on its consistency with the evidence. Our zero-shot framework outperforms fully-supervised approaches, as demonstrated by experiments on the FEVER and SciFact datasets, where our outputs are shown to be more faithful. More importantly, the decomposability nature of our framework inherently provides interpretability. Additionally, to reveal the most suitable metrics for evaluating factual error corrections, we analyze the correlation between commonly used metrics with human judgments in terms of three different dimensions regarding intelligibility and faithfulness. | [
"Huang, Kung-Hsiang",
"Chan, Hou Pong",
"Ji, Heng"
] | Zero-shot Faithful Factual Error Correction | acl-long.311 | Poster | 2305.07982 | [
"https://github.com/khuangaf/zerofec"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.312.bib | https://aclanthology.org/2023.acl-long.312/ | @inproceedings{li-etal-2023-open,
title = "Open-Domain Hierarchical Event Schema Induction by Incremental Prompting and Verification",
author = "Li, Sha and
Zhao, Ruining and
Li, Manling and
Ji, Heng and
Callison-Burch, Chris and
Han, Jiawei",
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.312",
doi = "10.18653/v1/2023.acl-long.312",
pages = "5677--5697",
abstract = "Event schemas are a form of world knowledge about the typical progression of events. Recent methods for event schema induction use information extraction systems to construct a large number of event graph instances from documents, and then learn to generalize the schema from such instances. In contrast, we propose to treat event schemas as a form of commonsense knowledge that can be derived from large language models (LLMs). This new paradigm greatly simplifies the schema induction process and allows us to handle both hierarchical relations and temporal relations between events in a straightforward way. Since event schemas have complex graph structures, we design an incremental prompting and verification method IncPrompt to break down the construction of a complex event graph into three stages: event skeleton construction, event expansion, and event-event relation verification. Compared to directly using LLMs to generate a linearized graph, IncSchema can generate large and complex schemas with 7.2{\%} F1 improvement in temporal relations and 31.0{\%} F1 improvement in hierarchical relations. In addition, compared to the previous state-of-the-art closed-domain schema induction model, human assessors were able to cover {\textasciitilde}10{\%} more events when translating the schemas into coherent stories and rated our schemas 1.3 points higher (on a 5-point scale) in terms of readability.",
}
| Event schemas are a form of world knowledge about the typical progression of events. Recent methods for event schema induction use information extraction systems to construct a large number of event graph instances from documents, and then learn to generalize the schema from such instances. In contrast, we propose to treat event schemas as a form of commonsense knowledge that can be derived from large language models (LLMs). This new paradigm greatly simplifies the schema induction process and allows us to handle both hierarchical relations and temporal relations between events in a straightforward way. Since event schemas have complex graph structures, we design an incremental prompting and verification method IncPrompt to break down the construction of a complex event graph into three stages: event skeleton construction, event expansion, and event-event relation verification. Compared to directly using LLMs to generate a linearized graph, IncSchema can generate large and complex schemas with 7.2{\%} F1 improvement in temporal relations and 31.0{\%} F1 improvement in hierarchical relations. In addition, compared to the previous state-of-the-art closed-domain schema induction model, human assessors were able to cover {\textasciitilde}10{\%} more events when translating the schemas into coherent stories and rated our schemas 1.3 points higher (on a 5-point scale) in terms of readability. | [
"Li, Sha",
"Zhao, Ruining",
"Li, Manling",
"Ji, Heng",
"Callison-Burch, Chris",
"Han, Jiawei"
] | Open-Domain Hierarchical Event Schema Induction by Incremental Prompting and Verification | acl-long.312 | Poster | 2307.01972 | [
"https://github.com/raspberryice/inc-schema"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.313.bib | https://aclanthology.org/2023.acl-long.313/ | @inproceedings{chakraborty-etal-2023-zero,
title = "Zero-shot Approach to Overcome Perturbation Sensitivity of Prompts",
author = "Chakraborty, Mohna and
Kulkarni, Adithya and
Li, Qi",
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.313",
doi = "10.18653/v1/2023.acl-long.313",
pages = "5698--5711",
abstract = "Recent studies have demonstrated that natural-language prompts can help to leverage the knowledge learned by pre-trained language models for the binary sentence-level sentiment classification task. Specifically, these methods utilize few-shot learning settings to fine-tune the sentiment classification model using manual or automatically generated prompts. However, the performance of these methods is sensitive to the perturbations of the utilized prompts. Furthermore, these methods depend on a few labeled instances for automatic prompt generation and prompt ranking. This study aims to find high-quality prompts for the given task in a zero-shot setting. Given a base prompt, our proposed approach automatically generates multiple prompts similar to the base prompt employing positional, reasoning, and paraphrasing techniques and then ranks the prompts using a novel metric. We empirically demonstrate that the top-ranked prompts are high-quality and significantly outperform the base prompt and the prompts generated using few-shot learning for the binary sentence-level sentiment classification task.",
}
| Recent studies have demonstrated that natural-language prompts can help to leverage the knowledge learned by pre-trained language models for the binary sentence-level sentiment classification task. Specifically, these methods utilize few-shot learning settings to fine-tune the sentiment classification model using manual or automatically generated prompts. However, the performance of these methods is sensitive to the perturbations of the utilized prompts. Furthermore, these methods depend on a few labeled instances for automatic prompt generation and prompt ranking. This study aims to find high-quality prompts for the given task in a zero-shot setting. Given a base prompt, our proposed approach automatically generates multiple prompts similar to the base prompt employing positional, reasoning, and paraphrasing techniques and then ranks the prompts using a novel metric. We empirically demonstrate that the top-ranked prompts are high-quality and significantly outperform the base prompt and the prompts generated using few-shot learning for the binary sentence-level sentiment classification task. | [
"Chakraborty, Mohna",
"Kulkarni, Adithya",
"Li, Qi"
] | Zero-shot Approach to Overcome Perturbation Sensitivity of Prompts | acl-long.313 | Poster | 2305.15689 | [
"https://github.com/mohna0310/zssc"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.314.bib | https://aclanthology.org/2023.acl-long.314/ | @inproceedings{schmidt-etal-2023-free,
title = "Free Lunch: Robust Cross-Lingual Transfer via Model Checkpoint Averaging",
author = "Schmidt, Fabian David and
Vuli{\'c}, Ivan and
Glava{\v{s}}, Goran",
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.314",
doi = "10.18653/v1/2023.acl-long.314",
pages = "5712--5730",
abstract = "Massively multilingual language models have displayed strong performance in zero-shot (ZS-XLT) and few-shot (FS-XLT) cross-lingual transfer setups, where models fine-tuned on task data in a source language are transferred without any or with only a few annotated instances to the target language(s). However, current work typically overestimates model performance as fine-tuned models are frequently evaluated at model checkpoints that generalize best to validation instances in the target languages. This effectively violates the main assumptions of {`}true{'} ZS-XLT and FS-XLT. Such XLT setups require robust methods that do not depend on labeled target language data for validation and model selection. In this work, aiming to improve the robustness of {`}true{'} ZS-XLT and FS-XLT, we propose a simple and effective method that averages different checkpoints (i.e., model snapshots) during task fine-tuning. We conduct exhaustive ZS-XLT and FS-XLT experiments across higher-level semantic tasks (NLI, extractive QA) and lower-level token classification tasks (NER, POS). The results indicate that averaging model checkpoints yields systematic and consistent performance gains across diverse target languages in all tasks. Importantly, it simultaneously substantially desensitizes XLT to varying hyperparameter choices in the absence of target language validation. We also show that checkpoint averaging benefits performance when further combined with run averaging (i.e., averaging the parameters of models fine-tuned over independent runs).",
}
| Massively multilingual language models have displayed strong performance in zero-shot (ZS-XLT) and few-shot (FS-XLT) cross-lingual transfer setups, where models fine-tuned on task data in a source language are transferred without any or with only a few annotated instances to the target language(s). However, current work typically overestimates model performance as fine-tuned models are frequently evaluated at model checkpoints that generalize best to validation instances in the target languages. This effectively violates the main assumptions of {`}true{'} ZS-XLT and FS-XLT. Such XLT setups require robust methods that do not depend on labeled target language data for validation and model selection. In this work, aiming to improve the robustness of {`}true{'} ZS-XLT and FS-XLT, we propose a simple and effective method that averages different checkpoints (i.e., model snapshots) during task fine-tuning. We conduct exhaustive ZS-XLT and FS-XLT experiments across higher-level semantic tasks (NLI, extractive QA) and lower-level token classification tasks (NER, POS). The results indicate that averaging model checkpoints yields systematic and consistent performance gains across diverse target languages in all tasks. Importantly, it simultaneously substantially desensitizes XLT to varying hyperparameter choices in the absence of target language validation. We also show that checkpoint averaging benefits performance when further combined with run averaging (i.e., averaging the parameters of models fine-tuned over independent runs). | [
"Schmidt, Fabian David",
"Vuli{\\'c}, Ivan",
"Glava{\\v{s}}, Goran"
] | Free Lunch: Robust Cross-Lingual Transfer via Model Checkpoint Averaging | acl-long.314 | Poster | 2305.16834 | [
"https://github.com/fdschmidt93/free-lunch-xlt"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.315.bib | https://aclanthology.org/2023.acl-long.315/ | @inproceedings{zeng-etal-2023-cross,
title = "Cross-View Language Modeling: Towards Unified Cross-Lingual Cross-Modal Pre-training",
author = "Zeng, Yan and
Zhou, Wangchunshu and
Luo, Ao and
Cheng, Ziming and
Zhang, Xinsong",
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.315",
doi = "10.18653/v1/2023.acl-long.315",
pages = "5731--5746",
abstract = "In this paper, we introduce Cross-View Language Modeling, a simple and effective pre-training framework that unifies cross-lingual and cross-modal pre-training with shared architectures and objectives. Our approach is motivated by a key observation that cross-lingual and cross-modal pre-training share the same goal of aligning two different views of the same object into a common semantic space. To this end, the cross-view language modeling framework considers both multi-modal data (i.e., image-caption pairs) and multi-lingual data (i.e., parallel sentence pairs) as two different views of the same object, and trains the model to align the two views by maximizing the mutual information between them with conditional masked language modeling and contrastive learning. We pre-train CCLM, a Cross-lingual Cross-modal Language Model, with the cross-view language modeling framework. Empirical results on IGLUE, a multi-lingual multi-modal benchmark, and two multi-lingual image-text retrieval datasets show that while conceptually simpler, CCLM significantly outperforms the prior state-of-the-art with an average absolute improvement of over 10{\%}. Moreover, CCLM is the first multi-lingual multi-modal pre-trained model that surpasses the translate-test performance of representative English vision-language models by zero-shot cross-lingual transfer.",
}
| In this paper, we introduce Cross-View Language Modeling, a simple and effective pre-training framework that unifies cross-lingual and cross-modal pre-training with shared architectures and objectives. Our approach is motivated by a key observation that cross-lingual and cross-modal pre-training share the same goal of aligning two different views of the same object into a common semantic space. To this end, the cross-view language modeling framework considers both multi-modal data (i.e., image-caption pairs) and multi-lingual data (i.e., parallel sentence pairs) as two different views of the same object, and trains the model to align the two views by maximizing the mutual information between them with conditional masked language modeling and contrastive learning. We pre-train CCLM, a Cross-lingual Cross-modal Language Model, with the cross-view language modeling framework. Empirical results on IGLUE, a multi-lingual multi-modal benchmark, and two multi-lingual image-text retrieval datasets show that while conceptually simpler, CCLM significantly outperforms the prior state-of-the-art with an average absolute improvement of over 10{\%}. Moreover, CCLM is the first multi-lingual multi-modal pre-trained model that surpasses the translate-test performance of representative English vision-language models by zero-shot cross-lingual transfer. | [
"Zeng, Yan",
"Zhou, Wangchunshu",
"Luo, Ao",
"Cheng, Ziming",
"Zhang, Xinsong"
] | Cross-View Language Modeling: Towards Unified Cross-Lingual Cross-Modal Pre-training | acl-long.315 | Poster | 2206.00621 | [
"https://github.com/zengyan-97/cclm"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.316.bib | https://aclanthology.org/2023.acl-long.316/ | @inproceedings{yang-etal-2023-unsupervised,
title = "Unsupervised Discontinuous Constituency Parsing with Mildly Context-Sensitive Grammars",
author = "Yang, Songlin and
Levy, Roger and
Kim, Yoon",
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.316",
doi = "10.18653/v1/2023.acl-long.316",
pages = "5747--5766",
abstract = "We study grammar induction with mildly context-sensitive grammars for unsupervised discontinuous parsing. Using the probabilistic linear context-free rewriting system (LCFRS) formalism, our approach fixes the rule structure in advance and focuses on parameter learning with maximum likelihood. To reduce the computational complexity of both parsing and parameter estimation, we restrict the grammar formalism to LCFRS-2 (i.e., binary LCFRS with fan-out two) and further discard rules that require $O(l^6)$ time to parse, reducing inference to $O(l^5)$. We find that using a large number of nonterminals is beneficial and thus make use of tensor decomposition-based rank-space dynamic programming with an embedding-based parameterization of rule probabilities to scale up the number of nonterminals. Experiments on German and Dutch show that our approach is able to induce linguistically meaningful trees with continuous and discontinuous structures.",
}
| We study grammar induction with mildly context-sensitive grammars for unsupervised discontinuous parsing. Using the probabilistic linear context-free rewriting system (LCFRS) formalism, our approach fixes the rule structure in advance and focuses on parameter learning with maximum likelihood. To reduce the computational complexity of both parsing and parameter estimation, we restrict the grammar formalism to LCFRS-2 (i.e., binary LCFRS with fan-out two) and further discard rules that require $O(l^6)$ time to parse, reducing inference to $O(l^5)$. We find that using a large number of nonterminals is beneficial and thus make use of tensor decomposition-based rank-space dynamic programming with an embedding-based parameterization of rule probabilities to scale up the number of nonterminals. Experiments on German and Dutch show that our approach is able to induce linguistically meaningful trees with continuous and discontinuous structures. | [
"Yang, Songlin",
"Levy, Roger",
"Kim, Yoon"
] | Unsupervised Discontinuous Constituency Parsing with Mildly Context-Sensitive Grammars | acl-long.316 | Oral | 2212.09140 | [
"https://github.com/sustcsonglin/tn-lcfrs"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.317.bib | https://aclanthology.org/2023.acl-long.317/ | @inproceedings{bhattamishra-etal-2023-simplicity,
title = "Simplicity Bias in Transformers and their Ability to Learn Sparse {B}oolean Functions",
author = "Bhattamishra, Satwik and
Patel, Arkil and
Kanade, Varun and
Blunsom, Phil",
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.317",
doi = "10.18653/v1/2023.acl-long.317",
pages = "5767--5791",
abstract = "Despite the widespread success of Transformers on NLP tasks, recent works have found that they struggle to model several formal languages when compared to recurrent models. This raises the question of why Transformers perform well in practice and whether they have any properties that enable them to generalize better than recurrent models. In this work, we conduct an extensive empirical study on Boolean functions to demonstrate the following: (i) Random Transformers are relatively more biased towards functions of low sensitivity. (ii) When trained on Boolean functions, both Transformers and LSTMs prioritize learning functions of low sensitivity, with Transformers ultimately converging to functions of lower sensitivity. (iii) On sparse Boolean functions which have low sensitivity, we find that Transformers generalize near perfectly even in the presence of noisy labels whereas LSTMs overfit and achieve poor generalization accuracy. Overall, our results provide strong quantifiable evidence that suggests differences in the inductive biases of Transformers and recurrent models which may help explain Transformer{'}s effective generalization performance despite relatively limited expressiveness.",
}
| Despite the widespread success of Transformers on NLP tasks, recent works have found that they struggle to model several formal languages when compared to recurrent models. This raises the question of why Transformers perform well in practice and whether they have any properties that enable them to generalize better than recurrent models. In this work, we conduct an extensive empirical study on Boolean functions to demonstrate the following: (i) Random Transformers are relatively more biased towards functions of low sensitivity. (ii) When trained on Boolean functions, both Transformers and LSTMs prioritize learning functions of low sensitivity, with Transformers ultimately converging to functions of lower sensitivity. (iii) On sparse Boolean functions which have low sensitivity, we find that Transformers generalize near perfectly even in the presence of noisy labels whereas LSTMs overfit and achieve poor generalization accuracy. Overall, our results provide strong quantifiable evidence that suggests differences in the inductive biases of Transformers and recurrent models which may help explain Transformer{'}s effective generalization performance despite relatively limited expressiveness. | [
"Bhattamishra, Satwik",
"Patel, Arkil",
"Kanade, Varun",
"Blunsom, Phil"
] | Simplicity Bias in Transformers and their Ability to Learn Sparse Boolean Functions | acl-long.317 | Poster | 2211.12316 | [
"https://github.com/satwik77/transformer-simplicity"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.318.bib | https://aclanthology.org/2023.acl-long.318/ | @inproceedings{gupta-etal-2023-counterspeeches,
title = "Counterspeeches up my sleeve! Intent Distribution Learning and Persistent Fusion for Intent-Conditioned Counterspeech Generation",
author = "Gupta, Rishabh and
Desai, Shaily and
Goel, Manvi and
Bandhakavi, Anil and
Chakraborty, Tanmoy and
Akhtar, Md. Shad",
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.318",
doi = "10.18653/v1/2023.acl-long.318",
pages = "5792--5809",
abstract = "Counterspeech has been demonstrated to be an efficacious approach for combating hate speech. While various conventional and controlled approaches have been studied in recent years to generate counterspeech, a counterspeech with a certain intent may not be sufficient in every scenario. Due to the complex and multifaceted nature of hate speech, utilizing multiple forms of counter-narratives with varying intents may be advantageous in different circumstances. In this paper, we explore intent-conditioned counterspeech generation. At first, we develop IntentCONAN, a diversified intent-specific counterspeech dataset with 6831 counterspeeches conditioned on five intents, i.e., informative, denouncing, question, positive, and humour. Subsequently, we propose QUARC, a two-stage framework for intent-conditioned counterspeech generation. QUARC leverages vector-quantized representations learned for each intent category along with PerFuMe, a novel fusion module to incorporate intent-specific information into the model. Our evaluation demonstrates that QUARC outperforms several baselines by an average of {\textasciitilde}10{\%} across evaluation metrics. An extensive human evaluation supplements our hypothesis of better and more appropriate responses than comparative systems.",
}
| Counterspeech has been demonstrated to be an efficacious approach for combating hate speech. While various conventional and controlled approaches have been studied in recent years to generate counterspeech, a counterspeech with a certain intent may not be sufficient in every scenario. Due to the complex and multifaceted nature of hate speech, utilizing multiple forms of counter-narratives with varying intents may be advantageous in different circumstances. In this paper, we explore intent-conditioned counterspeech generation. At first, we develop IntentCONAN, a diversified intent-specific counterspeech dataset with 6831 counterspeeches conditioned on five intents, i.e., informative, denouncing, question, positive, and humour. Subsequently, we propose QUARC, a two-stage framework for intent-conditioned counterspeech generation. QUARC leverages vector-quantized representations learned for each intent category along with PerFuMe, a novel fusion module to incorporate intent-specific information into the model. Our evaluation demonstrates that QUARC outperforms several baselines by an average of {\textasciitilde}10{\%} across evaluation metrics. An extensive human evaluation supplements our hypothesis of better and more appropriate responses than comparative systems. | [
"Gupta, Rishabh",
"Desai, Shaily",
"Goel, Manvi",
"B",
"hakavi, Anil",
"Chakraborty, Tanmoy",
"Akhtar, Md. Shad"
] | Counterspeeches up my sleeve! Intent Distribution Learning and Persistent Fusion for Intent-Conditioned Counterspeech Generation | acl-long.318 | Poster | 2305.13776 | [
"https://github.com/lcs2-iiitd/quarc-counterspeech"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.319.bib | https://aclanthology.org/2023.acl-long.319/ | @inproceedings{kothawade-etal-2023-ditto,
title = "{DITTO}: Data-efficient and Fair Targeted Subset Selection for {ASR} Accent Adaptation",
author = "Kothawade, Suraj and
Mekala, Anmol and
Hetha Havya, D.Chandra Sekhara and
Kothyari, Mayank and
Iyer, Rishabh and
Ramakrishnan, Ganesh and
Jyothi, Preethi",
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.319",
doi = "10.18653/v1/2023.acl-long.319",
pages = "5810--5822",
abstract = "State-of-the-art Automatic Speech Recognition (ASR) systems are known to exhibit disparate performance on varying speech accents. To improve performance on a specific target accent, a commonly adopted solution is to finetune the ASR model using accent-specific labeled speech. However, acquiring large amounts of labeled speech for specific target accents is challenging. Choosing an informative subset of speech samples that are most representative of the target accents becomes important for effective ASR finetuning. To address this problem, we propose DITTO (Data-efficient and faIr Targeted subseT selectiOn that uses Submodular Mutual Information (SMI) functions as acquisition functions to find the most informative set of utterances matching a target accent within a fixed budget. An important feature of DITTO is that it supports fair targeting for multiple accents, i.e. it can automatically select representative data points from multiple accents when the ASR model needs to perform well on more than one accent. We show that compared to other speech selection methods, DITTO is 3-5 times as label-efficient for its improvements on the Indic-TTS and L2 datasets.",
}
| State-of-the-art Automatic Speech Recognition (ASR) systems are known to exhibit disparate performance on varying speech accents. To improve performance on a specific target accent, a commonly adopted solution is to finetune the ASR model using accent-specific labeled speech. However, acquiring large amounts of labeled speech for specific target accents is challenging. Choosing an informative subset of speech samples that are most representative of the target accents becomes important for effective ASR finetuning. To address this problem, we propose DITTO (Data-efficient and faIr Targeted subseT selectiOn that uses Submodular Mutual Information (SMI) functions as acquisition functions to find the most informative set of utterances matching a target accent within a fixed budget. An important feature of DITTO is that it supports fair targeting for multiple accents, i.e. it can automatically select representative data points from multiple accents when the ASR model needs to perform well on more than one accent. We show that compared to other speech selection methods, DITTO is 3-5 times as label-efficient for its improvements on the Indic-TTS and L2 datasets. | [
"Kothawade, Suraj",
"Mekala, Anmol",
"Hetha Havya, D.Ch",
"ra Sekhara",
"Kothyari, Mayank",
"Iyer, Rishabh",
"Ramakrishnan, Ganesh",
"Jyothi, Preethi"
] | DITTO: Data-efficient and Fair Targeted Subset Selection for ASR Accent Adaptation | acl-long.319 | Poster | 2110.04908 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.320.bib | https://aclanthology.org/2023.acl-long.320/ | @inproceedings{zhao-etal-2023-verify,
title = "Verify-and-Edit: A Knowledge-Enhanced Chain-of-Thought Framework",
author = "Zhao, Ruochen and
Li, Xingxuan and
Joty, Shafiq and
Qin, Chengwei and
Bing, Lidong",
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.320",
doi = "10.18653/v1/2023.acl-long.320",
pages = "5823--5840",
abstract = "As large language models (LLMs) have become the norm in NLP, demonstrating good performance in generation and reasoning tasks, one of its most fatal disadvantages is the lack of factual correctness. Generating unfactual texts not only leads to lower performances but also degrades the trust and validity of their applications. Chain-of-Thought (CoT) prompting improves trust and model performance on complex reasoning tasks by generating interpretable reasoning chains, but still suffers from factuality concerns in knowledge-intensive tasks. In this paper, we propose the Verify-and-Edit framework for CoT prompting, which seeks to increase prediction factuality by post-editing reasoning chains according to external knowledge. Building on top of GPT-3, our framework lead to accuracy improvements in multiple open-domain question-answering tasks.",
}
| As large language models (LLMs) have become the norm in NLP, demonstrating good performance in generation and reasoning tasks, one of its most fatal disadvantages is the lack of factual correctness. Generating unfactual texts not only leads to lower performances but also degrades the trust and validity of their applications. Chain-of-Thought (CoT) prompting improves trust and model performance on complex reasoning tasks by generating interpretable reasoning chains, but still suffers from factuality concerns in knowledge-intensive tasks. In this paper, we propose the Verify-and-Edit framework for CoT prompting, which seeks to increase prediction factuality by post-editing reasoning chains according to external knowledge. Building on top of GPT-3, our framework lead to accuracy improvements in multiple open-domain question-answering tasks. | [
"Zhao, Ruochen",
"Li, Xingxuan",
"Joty, Shafiq",
"Qin, Chengwei",
"Bing, Lidong"
] | Verify-and-Edit: A Knowledge-Enhanced Chain-of-Thought Framework | acl-long.320 | Poster | 2305.03268 | [
"https://github.com/ruochenzhao/verify-and-edit"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.321.bib | https://aclanthology.org/2023.acl-long.321/ | @inproceedings{cao-etal-2023-bridging,
title = "Bridging the Domain Gaps in Context Representations for $k$-Nearest Neighbor Neural Machine Translation",
author = "Cao, Zhiwei and
Yang, Baosong and
Lin, Huan and
Wu, Suhang and
Wei, Xiangpeng and
Liu, Dayiheng and
Xie, Jun and
Zhang, Min and
Su, Jinsong",
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.321",
doi = "10.18653/v1/2023.acl-long.321",
pages = "5841--5853",
abstract = "$k$-Nearest neighbor machine translation ($k$NN-MT) has attracted increasing attention due to its ability to non-parametrically adapt to new translation domains. By using an upstream NMT model to traverse the downstream training corpus, it is equipped with a datastore containing vectorized key-value pairs, which are retrieved during inference to benefit translation.However, there often exists a significant gap between upstream and downstream domains, which hurts the datastore retrieval and the final translation quality.To deal with this issue, we propose a novel approach to boost the datastore retrieval of $k$NN-MT by reconstructing the original datastore.Concretely, we design a reviser to revise the key representations, making them better fit for the downstream domain. The reviser is trained using the collected semantically-related key-queries pairs, and optimized by two proposed losses: one is the key-queries semantic distance ensuring each revised key representation is semantically related to its corresponding queries, and the other is an L2-norm loss encouraging revised key representations to effectively retain the knowledge learned by the upstream NMT model. Extensive experiments on domain adaptation tasks demonstrate that our method can effectively boost the datastore retrieval and translation quality of $k$NN-MT.Our code is available at \url{https://github.com/DeepLearnXMU/Revised-knn-mt}.",
}
| $k$-Nearest neighbor machine translation ($k$NN-MT) has attracted increasing attention due to its ability to non-parametrically adapt to new translation domains. By using an upstream NMT model to traverse the downstream training corpus, it is equipped with a datastore containing vectorized key-value pairs, which are retrieved during inference to benefit translation.However, there often exists a significant gap between upstream and downstream domains, which hurts the datastore retrieval and the final translation quality.To deal with this issue, we propose a novel approach to boost the datastore retrieval of $k$NN-MT by reconstructing the original datastore.Concretely, we design a reviser to revise the key representations, making them better fit for the downstream domain. The reviser is trained using the collected semantically-related key-queries pairs, and optimized by two proposed losses: one is the key-queries semantic distance ensuring each revised key representation is semantically related to its corresponding queries, and the other is an L2-norm loss encouraging revised key representations to effectively retain the knowledge learned by the upstream NMT model. Extensive experiments on domain adaptation tasks demonstrate that our method can effectively boost the datastore retrieval and translation quality of $k$NN-MT.Our code is available at \url{https://github.com/DeepLearnXMU/Revised-knn-mt}. | [
"Cao, Zhiwei",
"Yang, Baosong",
"Lin, Huan",
"Wu, Suhang",
"Wei, Xiangpeng",
"Liu, Dayiheng",
"Xie, Jun",
"Zhang, Min",
"Su, Jinsong"
] | Bridging the Domain Gaps in Context Representations for k-Nearest Neighbor Neural Machine Translation | acl-long.321 | Poster | 2305.16599 | [
"https://github.com/deeplearnxmu/revisedkey-knn-mt"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.322.bib | https://aclanthology.org/2023.acl-long.322/ | @inproceedings{jundi-etal-2023-node,
title = "Node Placement in Argument Maps: Modeling Unidirectional Relations in High {\&} Low-Resource Scenarios",
author = "Jundi, Iman and
Falk, Neele and
Vecchi, Eva Maria and
Lapesa, Gabriella",
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.322",
doi = "10.18653/v1/2023.acl-long.322",
pages = "5854--5876",
abstract = "Argument maps structure discourse into nodes in a tree with each node being an argument that supports or opposes its parent argument. This format is more comprehensible and less redundant compared to an unstructured one. Exploring those maps and maintaining their structure by placing new arguments under suitable parents is more challenging for users with huge maps that are typical in online discussions. To support those users, we introduce the task of node placement: suggesting candidate nodes as parents for a new contribution. We establish an upper-bound of human performance, and conduct experiments with models of various sizes and training strategies. We experiment with a selection of maps from Kialo, drawn from a heterogeneous set of domains. Based on an annotation study, we highlight the ambiguity of the task that makes it challenging for both humans and models. We examine the unidirectional relation between tree nodes and show that encoding a node into different embeddings for each of the parent and child cases improves performance. We further show the few-shot effectiveness of our approach.",
}
| Argument maps structure discourse into nodes in a tree with each node being an argument that supports or opposes its parent argument. This format is more comprehensible and less redundant compared to an unstructured one. Exploring those maps and maintaining their structure by placing new arguments under suitable parents is more challenging for users with huge maps that are typical in online discussions. To support those users, we introduce the task of node placement: suggesting candidate nodes as parents for a new contribution. We establish an upper-bound of human performance, and conduct experiments with models of various sizes and training strategies. We experiment with a selection of maps from Kialo, drawn from a heterogeneous set of domains. Based on an annotation study, we highlight the ambiguity of the task that makes it challenging for both humans and models. We examine the unidirectional relation between tree nodes and show that encoding a node into different embeddings for each of the parent and child cases improves performance. We further show the few-shot effectiveness of our approach. | [
"Jundi, Iman",
"Falk, Neele",
"Vecchi, Eva Maria",
"Lapesa, Gabriella"
] | Node Placement in Argument Maps: Modeling Unidirectional Relations in High & Low-Resource Scenarios | acl-long.322 | Oral | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.323.bib | https://aclanthology.org/2023.acl-long.323/ | @inproceedings{philippy-etal-2023-towards,
title = "Towards a Common Understanding of Contributing Factors for Cross-Lingual Transfer in Multilingual Language Models: A Review",
author = "Philippy, Fred and
Guo, Siwen and
Haddadan, Shohreh",
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.323",
doi = "10.18653/v1/2023.acl-long.323",
pages = "5877--5891",
abstract = "In recent years, pre-trained Multilingual Language Models (MLLMs) have shown a strong ability to transfer knowledge across different languages. However, given that the aspiration for such an ability has not been explicitly incorporated in the design of the majority of MLLMs, it is challenging to obtain a unique and straightforward explanation for its emergence. In this review paper, we survey literature that investigates different factors contributing to the capacity of MLLMs to perform zero-shot cross-lingual transfer and subsequently outline and discuss these factors in detail. To enhance the structure of this review and to facilitate consolidation with future studies, we identify five categories of such factors. In addition to providing a summary of empirical evidence from past studies, we identify consensuses among studies with consistent findings and resolve conflicts among contradictory ones. Our work contextualizes and unifies existing research streams which aim at explaining the cross-lingual potential of MLLMs. This review provides, first, an aligned reference point for future research and, second, guidance for a better-informed and more efficient way of leveraging the cross-lingual capacity of MLLMs.",
}
| In recent years, pre-trained Multilingual Language Models (MLLMs) have shown a strong ability to transfer knowledge across different languages. However, given that the aspiration for such an ability has not been explicitly incorporated in the design of the majority of MLLMs, it is challenging to obtain a unique and straightforward explanation for its emergence. In this review paper, we survey literature that investigates different factors contributing to the capacity of MLLMs to perform zero-shot cross-lingual transfer and subsequently outline and discuss these factors in detail. To enhance the structure of this review and to facilitate consolidation with future studies, we identify five categories of such factors. In addition to providing a summary of empirical evidence from past studies, we identify consensuses among studies with consistent findings and resolve conflicts among contradictory ones. Our work contextualizes and unifies existing research streams which aim at explaining the cross-lingual potential of MLLMs. This review provides, first, an aligned reference point for future research and, second, guidance for a better-informed and more efficient way of leveraging the cross-lingual capacity of MLLMs. | [
"Philippy, Fred",
"Guo, Siwen",
"Haddadan, Shohreh"
] | Towards a Common Understanding of Contributing Factors for Cross-Lingual Transfer in Multilingual Language Models: A Review | acl-long.323 | Poster | 2305.16768 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.324.bib | https://aclanthology.org/2023.acl-long.324/ | @inproceedings{lu-etal-2023-toward,
title = "Toward Human-Like Evaluation for Natural Language Generation with Error Analysis",
author = "Lu, Qingyu and
Ding, Liang and
Xie, Liping and
Zhang, Kanjian and
Wong, Derek F. 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.324",
doi = "10.18653/v1/2023.acl-long.324",
pages = "5892--5907",
abstract = "The pretrained language model (PLM) based metrics have been successfully used in evaluating language generation tasks. Recent studies of the human evaluation community show that considering both major errors (e.g. mistranslated tokens) and minor errors (e.g. imperfections in fluency) can produce high-quality judgments. This inspires us to approach the final goal of the automatic metrics (human-like evaluations) by fine-grained error analysis. In this paper, we argue that the ability to estimate sentence confidence is the tip of the iceberg for PLM-based metrics. And it can be used to refine the generated sentence toward higher confidence and more reference-grounded, where the costs of refining and approaching reference are used to determine the major and minor errors, respectively. To this end, we take BARTScore as the testbed and present an innovative solution to marry the unexploited sentence refining capacity of BARTScore and human-like error analysis, where the final score consists of both the evaluations of major and minor errors. Experiments show that our solution consistently and significantly improves BARTScore, and outperforms top-scoring metrics in 19/25 test settings. Analyses demonstrate our method robustly and efficiently approaches human-like evaluations, enjoying better interpretability. Our code and scripts will be publicly released in \url{https://github.com/Coldmist-Lu/ErrorAnalysis_NLGEvaluation}.",
}
| The pretrained language model (PLM) based metrics have been successfully used in evaluating language generation tasks. Recent studies of the human evaluation community show that considering both major errors (e.g. mistranslated tokens) and minor errors (e.g. imperfections in fluency) can produce high-quality judgments. This inspires us to approach the final goal of the automatic metrics (human-like evaluations) by fine-grained error analysis. In this paper, we argue that the ability to estimate sentence confidence is the tip of the iceberg for PLM-based metrics. And it can be used to refine the generated sentence toward higher confidence and more reference-grounded, where the costs of refining and approaching reference are used to determine the major and minor errors, respectively. To this end, we take BARTScore as the testbed and present an innovative solution to marry the unexploited sentence refining capacity of BARTScore and human-like error analysis, where the final score consists of both the evaluations of major and minor errors. Experiments show that our solution consistently and significantly improves BARTScore, and outperforms top-scoring metrics in 19/25 test settings. Analyses demonstrate our method robustly and efficiently approaches human-like evaluations, enjoying better interpretability. Our code and scripts will be publicly released in \url{https://github.com/Coldmist-Lu/ErrorAnalysis_NLGEvaluation}. | [
"Lu, Qingyu",
"Ding, Liang",
"Xie, Liping",
"Zhang, Kanjian",
"Wong, Derek F.",
"Tao, Dacheng"
] | Toward Human-Like Evaluation for Natural Language Generation with Error Analysis | acl-long.324 | Oral | 2212.10179 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.325.bib | https://aclanthology.org/2023.acl-long.325/ | @inproceedings{wu-etal-2023-connective,
title = "Connective Prediction for Implicit Discourse Relation Recognition via Knowledge Distillation",
author = "Wu, Hongyi and
Zhou, Hao and
Lan, Man and
Wu, Yuanbin and
Zhang, Yadong",
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.325",
doi = "10.18653/v1/2023.acl-long.325",
pages = "5908--5923",
abstract = "Implicit discourse relation recognition (IDRR) remains a challenging task in discourse analysis due to the absence of connectives. Most existing methods utilize one-hot labels as the sole optimization target, ignoring the internal association among connectives. Besides, these approaches spend lots of effort on template construction, negatively affecting the generalization capability. To address these problems,we propose a novel Connective Prediction via Knowledge Distillation (CP-KD) approach to instruct large-scale pre-trained language models (PLMs) mining the latent correlations between connectives and discourse relations, which is meaningful for IDRR. Experimental results on the PDTB 2.0/3.0 and CoNLL2016 datasets show that our method significantly outperforms the state-of-the-art models on coarse-grained and fine-grained discourse relations. Moreover, our approach can be transferred to explicit discourse relation recognition(EDRR) and achieve acceptable performance.",
}
| Implicit discourse relation recognition (IDRR) remains a challenging task in discourse analysis due to the absence of connectives. Most existing methods utilize one-hot labels as the sole optimization target, ignoring the internal association among connectives. Besides, these approaches spend lots of effort on template construction, negatively affecting the generalization capability. To address these problems,we propose a novel Connective Prediction via Knowledge Distillation (CP-KD) approach to instruct large-scale pre-trained language models (PLMs) mining the latent correlations between connectives and discourse relations, which is meaningful for IDRR. Experimental results on the PDTB 2.0/3.0 and CoNLL2016 datasets show that our method significantly outperforms the state-of-the-art models on coarse-grained and fine-grained discourse relations. Moreover, our approach can be transferred to explicit discourse relation recognition(EDRR) and achieve acceptable performance. | [
"Wu, Hongyi",
"Zhou, Hao",
"Lan, Man",
"Wu, Yuanbin",
"Zhang, Yadong"
] | Connective Prediction for Implicit Discourse Relation Recognition via Knowledge Distillation | acl-long.325 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.326.bib | https://aclanthology.org/2023.acl-long.326/ | @inproceedings{cao-2023-best,
title = "What is the best recipe for character-level encoder-only modelling?",
author = "Cao, Kris",
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.326",
doi = "10.18653/v1/2023.acl-long.326",
pages = "5924--5938",
abstract = "This paper aims to benchmark recent progress in language understanding models that output contextualised representations at the character level. Many such modelling architectures and methods to train those architectures have been proposed, but it is currently unclear what the relative contributions of the architecture vs. the pretraining objective are to final model performance. We explore the design space of such models, comparing architectural innovations (Clark et al., 2022, Jaegle et al., 2022, Tay et al., 2021) and a variety of different pretraining objectives on a suite of evaluation tasks with a fixed training procedure in order to find the currently optimal way to build and train character-level BERT-like models. We find that our best performing character-level model exceeds the performance of a token-based model trained with the same settings on the same data, suggesting that character-level models are ready for more widespread adoption. Unfortunately, the best method to train character-level models still relies on a subword-level tokeniser during pretraining, and final model performance is highly dependent on tokeniser quality. We believe our results demonstrate the readiness of character-level models for multilingual language representation, and encourage NLP practitioners to try them as drop-in replacements for token-based models.",
}
| This paper aims to benchmark recent progress in language understanding models that output contextualised representations at the character level. Many such modelling architectures and methods to train those architectures have been proposed, but it is currently unclear what the relative contributions of the architecture vs. the pretraining objective are to final model performance. We explore the design space of such models, comparing architectural innovations (Clark et al., 2022, Jaegle et al., 2022, Tay et al., 2021) and a variety of different pretraining objectives on a suite of evaluation tasks with a fixed training procedure in order to find the currently optimal way to build and train character-level BERT-like models. We find that our best performing character-level model exceeds the performance of a token-based model trained with the same settings on the same data, suggesting that character-level models are ready for more widespread adoption. Unfortunately, the best method to train character-level models still relies on a subword-level tokeniser during pretraining, and final model performance is highly dependent on tokeniser quality. We believe our results demonstrate the readiness of character-level models for multilingual language representation, and encourage NLP practitioners to try them as drop-in replacements for token-based models. | [
"Cao, Kris"
] | What is the best recipe for character-level encoder-only modelling? | acl-long.326 | Oral | 2305.05461 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.327.bib | https://aclanthology.org/2023.acl-long.327/ | @inproceedings{li-etal-2023-unifying,
title = "Unifying Cross-Lingual and Cross-Modal Modeling Towards Weakly Supervised Multilingual Vision-Language Pre-training",
author = "Li, Zejun and
Fan, Zhihao and
Chen, Jingjing and
Zhang, Qi and
Huang, Xuanjing and
Wei, Zhongyu",
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.327",
doi = "10.18653/v1/2023.acl-long.327",
pages = "5939--5958",
abstract = "Multilingual Vision-Language Pre-training (VLP) is a promising but challenging topic due to the lack of large-scale multilingual image-text pairs. Existing works address the problem by translating English data into other languages, which is intuitive and the generated data is usually limited in form and scale. In this paper, we explore a more practical and scalable setting: weakly supervised multilingual VLP with only English image-text pairs and multilingual text corpora. We argue that the universal multilingual representation learned from texts allows the cross-modal interaction learned in English to be transferable to other languages. To this end, we propose a framework to effectively unify cross-lingual and cross-modal pre-training. For unified modeling on different data, we design an architecture with flexible modules to learn different interactions. Moreover, two unified tasks are introduced to efficiently guide the unified cross-lingual cross-modal learning. Extensive experiments demonstrate that our pre-trained model learns universal multilingual multimodal representations, allowing effective cross-lingual transfer on multimodal tasks. Code and models are available at \url{https://github.com/FudanDISC/weakly-supervised-mVLP}.",
}
| Multilingual Vision-Language Pre-training (VLP) is a promising but challenging topic due to the lack of large-scale multilingual image-text pairs. Existing works address the problem by translating English data into other languages, which is intuitive and the generated data is usually limited in form and scale. In this paper, we explore a more practical and scalable setting: weakly supervised multilingual VLP with only English image-text pairs and multilingual text corpora. We argue that the universal multilingual representation learned from texts allows the cross-modal interaction learned in English to be transferable to other languages. To this end, we propose a framework to effectively unify cross-lingual and cross-modal pre-training. For unified modeling on different data, we design an architecture with flexible modules to learn different interactions. Moreover, two unified tasks are introduced to efficiently guide the unified cross-lingual cross-modal learning. Extensive experiments demonstrate that our pre-trained model learns universal multilingual multimodal representations, allowing effective cross-lingual transfer on multimodal tasks. Code and models are available at \url{https://github.com/FudanDISC/weakly-supervised-mVLP}. | [
"Li, Zejun",
"Fan, Zhihao",
"Chen, Jingjing",
"Zhang, Qi",
"Huang, Xuanjing",
"Wei, Zhongyu"
] | Unifying Cross-Lingual and Cross-Modal Modeling Towards Weakly Supervised Multilingual Vision-Language Pre-training | acl-long.327 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.328.bib | https://aclanthology.org/2023.acl-long.328/ | @inproceedings{ma-etal-2023-learning,
title = "Learning {``}{O}{''} Helps for Learning More: Handling the Unlabeled Entity Problem for Class-incremental {NER}",
author = "Ma, Ruotian and
Chen, Xuanting and
Lin, Zhang and
Zhou, Xin and
Wang, Junzhe and
Gui, Tao and
Zhang, Qi and
Gao, Xiang and
Chen, Yun Wen",
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.328",
doi = "10.18653/v1/2023.acl-long.328",
pages = "5959--5979",
abstract = "As the categories of named entities rapidly increase, the deployed NER models are required to keep updating toward recognizing more entity types, creating a demand for class-incremental learning for NER. Considering the privacy concerns and storage constraints, the standard paradigm for class-incremental NER updates the models with training data only annotated with the new classes, yet the entities from other entity classes are regarded as {``}Non-entity{''} (or {``}O{''}). In this work, we conduct an empirical study on the {``}Unlabeled Entity Problem{''} and find that it leads to severe confusion between {``}O{''} and entities, decreasing class discrimination of old classes and declining the model{'}s ability to learn new classes. To solve the Unlabeled Entity Problem, we propose a novel representation learning method to learn discriminative representations for the entity classes and {``}O{''}. Specifically, we propose an entity-aware contrastive learning method that adaptively detects entity clusters in {``}O{''}. Furthermore, we propose two effective distance-based relabeling strategies for better learning the old classes. We introduce a more realistic and challenging benchmark for class-incremental NER, and the proposed method achieves up to 10.62{\%} improvement over the baseline methods.",
}
| As the categories of named entities rapidly increase, the deployed NER models are required to keep updating toward recognizing more entity types, creating a demand for class-incremental learning for NER. Considering the privacy concerns and storage constraints, the standard paradigm for class-incremental NER updates the models with training data only annotated with the new classes, yet the entities from other entity classes are regarded as {``}Non-entity{''} (or {``}O{''}). In this work, we conduct an empirical study on the {``}Unlabeled Entity Problem{''} and find that it leads to severe confusion between {``}O{''} and entities, decreasing class discrimination of old classes and declining the model{'}s ability to learn new classes. To solve the Unlabeled Entity Problem, we propose a novel representation learning method to learn discriminative representations for the entity classes and {``}O{''}. Specifically, we propose an entity-aware contrastive learning method that adaptively detects entity clusters in {``}O{''}. Furthermore, we propose two effective distance-based relabeling strategies for better learning the old classes. We introduce a more realistic and challenging benchmark for class-incremental NER, and the proposed method achieves up to 10.62{\%} improvement over the baseline methods. | [
"Ma, Ruotian",
"Chen, Xuanting",
"Lin, Zhang",
"Zhou, Xin",
"Wang, Junzhe",
"Gui, Tao",
"Zhang, Qi",
"Gao, Xiang",
"Chen, Yun Wen"
] | Learning “O” Helps for Learning More: Handling the Unlabeled Entity Problem for Class-incremental NER | acl-long.328 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.329.bib | https://aclanthology.org/2023.acl-long.329/ | @inproceedings{fei-etal-2023-scene,
title = "Scene Graph as Pivoting: Inference-time Image-free Unsupervised Multimodal Machine Translation with Visual Scene Hallucination",
author = "Fei, Hao and
Liu, Qian and
Zhang, Meishan and
Zhang, Min and
Chua, Tat-Seng",
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.329",
doi = "10.18653/v1/2023.acl-long.329",
pages = "5980--5994",
abstract = "In this work, we investigate a more realistic unsupervised multimodal machine translation (UMMT) setup, inference-time image-free UMMT, where the model is trained with source-text image pairs, and tested with only source-text inputs. First, we represent the input images and texts with the visual and language scene graphs (SG), where such fine-grained vision-language features ensure a holistic understanding of the semantics. To enable pure-text input during inference, we devise a visual scene hallucination mechanism that dynamically generates pseudo visual SG from the given textual SG. Several SG-pivoting based learning objectives are introduced for unsupervised translation training. On the benchmark Multi30K data, our SG-based method outperforms the best-performing baseline by significant BLEU scores on the task and setup, helping yield translations with better completeness, relevance and fluency without relying on paired images. Further in-depth analyses reveal how our model advances in the task setting.",
}
| In this work, we investigate a more realistic unsupervised multimodal machine translation (UMMT) setup, inference-time image-free UMMT, where the model is trained with source-text image pairs, and tested with only source-text inputs. First, we represent the input images and texts with the visual and language scene graphs (SG), where such fine-grained vision-language features ensure a holistic understanding of the semantics. To enable pure-text input during inference, we devise a visual scene hallucination mechanism that dynamically generates pseudo visual SG from the given textual SG. Several SG-pivoting based learning objectives are introduced for unsupervised translation training. On the benchmark Multi30K data, our SG-based method outperforms the best-performing baseline by significant BLEU scores on the task and setup, helping yield translations with better completeness, relevance and fluency without relying on paired images. Further in-depth analyses reveal how our model advances in the task setting. | [
"Fei, Hao",
"Liu, Qian",
"Zhang, Meishan",
"Zhang, Min",
"Chua, Tat-Seng"
] | Scene Graph as Pivoting: Inference-time Image-free Unsupervised Multimodal Machine Translation with Visual Scene Hallucination | acl-long.329 | Poster | 2305.12256 | [
"https://github.com/scofield7419/ummt-vsh"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.330.bib | https://aclanthology.org/2023.acl-long.330/ | @inproceedings{ma-etal-2023-colada,
title = "{C}o{L}a{D}a: A Collaborative Label Denoising Framework for Cross-lingual Named Entity Recognition",
author = {Ma, Tingting and
Wu, Qianhui and
Jiang, Huiqiang and
Karlsson, B{\"o}rje and
Zhao, Tiejun and
Lin, Chin-Yew},
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.330",
doi = "10.18653/v1/2023.acl-long.330",
pages = "5995--6009",
abstract = "Cross-lingual named entity recognition (NER) aims to train an NER system that generalizes well to a target language by leveraging labeled data in a given source language. Previous work alleviates the data scarcity problem by translating source-language labeled data or performing knowledge distillation on target-language unlabeled data. However, these methods may suffer from label noise due to the automatic labeling process. In this paper, we propose CoLaDa, a Collaborative Label Denoising Framework, to address this problem. Specifically, we first explore a model-collaboration-based denoising scheme that enables models trained on different data sources to collaboratively denoise pseudo labels used by each other. We then present an instance-collaboration-based strategy that considers the label consistency of each token{'}s neighborhood in the representation space for denoising. Experiments on different benchmark datasets show that the proposed CoLaDa achieves superior results compared to previous methods, especially when generalizing to distant languages.",
}
| Cross-lingual named entity recognition (NER) aims to train an NER system that generalizes well to a target language by leveraging labeled data in a given source language. Previous work alleviates the data scarcity problem by translating source-language labeled data or performing knowledge distillation on target-language unlabeled data. However, these methods may suffer from label noise due to the automatic labeling process. In this paper, we propose CoLaDa, a Collaborative Label Denoising Framework, to address this problem. Specifically, we first explore a model-collaboration-based denoising scheme that enables models trained on different data sources to collaboratively denoise pseudo labels used by each other. We then present an instance-collaboration-based strategy that considers the label consistency of each token{'}s neighborhood in the representation space for denoising. Experiments on different benchmark datasets show that the proposed CoLaDa achieves superior results compared to previous methods, especially when generalizing to distant languages. | [
"Ma, Tingting",
"Wu, Qianhui",
"Jiang, Huiqiang",
"Karlsson, B{\\\"o}rje",
"Zhao, Tiejun",
"Lin, Chin-Yew"
] | CoLaDa: A Collaborative Label Denoising Framework for Cross-lingual Named Entity Recognition | acl-long.330 | Poster | 2305.14913 | [
"https://github.com/microsoft/vert-papers"
] | https://huggingface.co/papers/2305.14913 | 2 | 1 | 0 | 6 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.331.bib | https://aclanthology.org/2023.acl-long.331/ | @inproceedings{sun-etal-2023-dialect,
title = "Dialect-robust Evaluation of Generated Text",
author = "Sun, Jiao and
Sellam, Thibault and
Clark, Elizabeth and
Vu, Tu and
Dozat, Timothy and
Garrette, Dan and
Siddhant, Aditya and
Eisenstein, Jacob and
Gehrmann, Sebastian",
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.331",
doi = "10.18653/v1/2023.acl-long.331",
pages = "6010--6028",
abstract = "Text generation metrics that are not robust to dialect variation make it impossible to tell how well systems perform for many groups of users, and can even penalize systems for producing text in lower-resource dialects. In this paper, we introduce a suite of methods to assess whether metrics are dialect robust. These methods show that state-of-the-art metrics are not dialect robust: they often prioritize dialect similarity over semantics, preferring outputs that are semantically incorrect over outputs that match the semantics of the reference but contain dialect differences. As a step towards dialect-robust metrics for text generation, we propose NANO, which introduces regional and language information to the metric{'}s pretraining. NANO significantly improves dialect robustness while preserving the correlation between automated metrics and human ratings. It also enables a more ambitious approach to evaluation, dialect awareness, in which system outputs are scored by both semantic match to the reference and appropriateness in any specified dialect.",
}
| Text generation metrics that are not robust to dialect variation make it impossible to tell how well systems perform for many groups of users, and can even penalize systems for producing text in lower-resource dialects. In this paper, we introduce a suite of methods to assess whether metrics are dialect robust. These methods show that state-of-the-art metrics are not dialect robust: they often prioritize dialect similarity over semantics, preferring outputs that are semantically incorrect over outputs that match the semantics of the reference but contain dialect differences. As a step towards dialect-robust metrics for text generation, we propose NANO, which introduces regional and language information to the metric{'}s pretraining. NANO significantly improves dialect robustness while preserving the correlation between automated metrics and human ratings. It also enables a more ambitious approach to evaluation, dialect awareness, in which system outputs are scored by both semantic match to the reference and appropriateness in any specified dialect. | [
"Sun, Jiao",
"Sellam, Thibault",
"Clark, Elizabeth",
"Vu, Tu",
"Dozat, Timothy",
"Garrette, Dan",
"Siddhant, Aditya",
"Eisenstein, Jacob",
"Gehrmann, Sebastian"
] | Dialect-robust Evaluation of Generated Text | acl-long.331 | Poster | 2211.00922 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.332.bib | https://aclanthology.org/2023.acl-long.332/ | @inproceedings{zhang-etal-2023-understanding,
title = "Understanding and Improving the Robustness of Terminology Constraints in Neural Machine Translation",
author = "Zhang, Huaao and
Wang, Qiang and
Qin, Bo and
Shi, Zelin and
Wang, Haibo and
Chen, Ming",
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.332",
doi = "10.18653/v1/2023.acl-long.332",
pages = "6029--6042",
abstract = "In this work, we study the robustness of two typical terminology translation methods: Placeholder (PH) and Code-Switch (CS), concerning (1) the number of constraints and (2) the target constraint length. We identify that existing terminology constraint test sets, such as IATE, Wiktionary, and TICO, are blind to this issue due to oversimplified constraint settings. To solve it, we create a new challenging test set of English-German, increasing the average constraint count per sentence from 1.1{\textasciitilde}1.7 to 6.1 and the length per target constraint from 1.1{\textasciitilde}1.2 words to 3.4 words. Then we find that PH and CS methods degrade as the number of constraints increases, but they have complementary strengths. Specifically, PH is better at retaining high constraint accuracy but lower translation quality as measured by BLEU and COMET scores. In contrast, CS has the opposite results. Based on these observations, we propose a simple but effective method combining the advantages of PH and CS. This approach involves training a model like PH to predict the term labels, and then during inference replacing those labels with target terminology text like CS, so that the subsequent generation is aware of the target term content. Extensive experimental results show that this approach can achieve high constraint accuracy and translation quality simultaneously, regardless of the number or length of constraints.",
}
| In this work, we study the robustness of two typical terminology translation methods: Placeholder (PH) and Code-Switch (CS), concerning (1) the number of constraints and (2) the target constraint length. We identify that existing terminology constraint test sets, such as IATE, Wiktionary, and TICO, are blind to this issue due to oversimplified constraint settings. To solve it, we create a new challenging test set of English-German, increasing the average constraint count per sentence from 1.1{\textasciitilde}1.7 to 6.1 and the length per target constraint from 1.1{\textasciitilde}1.2 words to 3.4 words. Then we find that PH and CS methods degrade as the number of constraints increases, but they have complementary strengths. Specifically, PH is better at retaining high constraint accuracy but lower translation quality as measured by BLEU and COMET scores. In contrast, CS has the opposite results. Based on these observations, we propose a simple but effective method combining the advantages of PH and CS. This approach involves training a model like PH to predict the term labels, and then during inference replacing those labels with target terminology text like CS, so that the subsequent generation is aware of the target term content. Extensive experimental results show that this approach can achieve high constraint accuracy and translation quality simultaneously, regardless of the number or length of constraints. | [
"Zhang, Huaao",
"Wang, Qiang",
"Qin, Bo",
"Shi, Zelin",
"Wang, Haibo",
"Chen, Ming"
] | Understanding and Improving the Robustness of Terminology Constraints in Neural Machine Translation | acl-long.332 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.333.bib | https://aclanthology.org/2023.acl-long.333/ | @inproceedings{sinha-etal-2023-language,
title = "Language model acceptability judgements are not always robust to context",
author = "Sinha, Koustuv and
Gauthier, Jon and
Mueller, Aaron and
Misra, Kanishka and
Fuentes, Keren and
Levy, Roger and
Williams, Adina",
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.333",
doi = "10.18653/v1/2023.acl-long.333",
pages = "6043--6063",
abstract = "Targeted syntactic evaluations of language models ask whether models show stable preferences for syntactically acceptable content over minimal-pair unacceptable inputs. Our best syntactic evaluation datasets, however, provide substantially less linguistic context than models receive during pretraining. This mismatch raises an important question: how robust are models{'} syntactic judgements across different contexts? In this paper, we vary the input contexts based on: length, the types of syntactic phenomena it contains, and whether or not there are grammatical violations. We find that model judgements are generally robust when placed in randomly sampled linguistic contexts, but are unstable when contexts match the test stimuli in syntactic structure. Among all tested models (GPT-2 and five variants of OPT), we find that model performance is affected when we provided contexts with matching syntactic structure: performance significantly improves when contexts are acceptable, and it significantly declines when they are unacceptable. This effect is amplified by the length of the context, except for unrelated inputs. We show that these changes in model performance are not explainable by acceptability-preserving syntactic perturbations. This sensitivity to highly specific syntactic features of the context can only be explained by the models{'} implicit in-context learning abilities.",
}
| Targeted syntactic evaluations of language models ask whether models show stable preferences for syntactically acceptable content over minimal-pair unacceptable inputs. Our best syntactic evaluation datasets, however, provide substantially less linguistic context than models receive during pretraining. This mismatch raises an important question: how robust are models{'} syntactic judgements across different contexts? In this paper, we vary the input contexts based on: length, the types of syntactic phenomena it contains, and whether or not there are grammatical violations. We find that model judgements are generally robust when placed in randomly sampled linguistic contexts, but are unstable when contexts match the test stimuli in syntactic structure. Among all tested models (GPT-2 and five variants of OPT), we find that model performance is affected when we provided contexts with matching syntactic structure: performance significantly improves when contexts are acceptable, and it significantly declines when they are unacceptable. This effect is amplified by the length of the context, except for unrelated inputs. We show that these changes in model performance are not explainable by acceptability-preserving syntactic perturbations. This sensitivity to highly specific syntactic features of the context can only be explained by the models{'} implicit in-context learning abilities. | [
"Sinha, Koustuv",
"Gauthier, Jon",
"Mueller, Aaron",
"Misra, Kanishka",
"Fuentes, Keren",
"Levy, Roger",
"Williams, Adina"
] | Language model acceptability judgements are not always robust to context | acl-long.333 | Poster | 2212.08979 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.334.bib | https://aclanthology.org/2023.acl-long.334/ | @inproceedings{zhao-etal-2023-robut,
title = "{R}obu{T}: A Systematic Study of Table {QA} Robustness Against Human-Annotated Adversarial Perturbations",
author = "Zhao, Yilun and
Zhao, Chen and
Nan, Linyong and
Qi, Zhenting and
Zhang, Wenlin and
Tang, Xiangru and
Mi, Boyu and
Radev, Dragomir",
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.334",
doi = "10.18653/v1/2023.acl-long.334",
pages = "6064--6081",
abstract = "Despite significant progress having been made in question answering on tabular data (Table QA), it{'}s unclear whether, and to what extent existing Table QA models are robust to task-specific perturbations, e.g., replacing key question entities or shuffling table columns. To systematically study the robustness of Table QA models, we propose a benchmark called RobuT, which builds upon existing Table QA datasets (WTQ, WikiSQL-Weak, and SQA) and includes human-annotated adversarial perturbations in terms of table header, table content, and question. Our results indicate that both state-of-the-art Table QA models and large language models (e.g., GPT-3) with few-shot learning falter in these adversarial sets. We propose to address this problem by using large language models to generate adversarial examples to enhance training, which significantly improves the robustness of Table QA models.",
}
| Despite significant progress having been made in question answering on tabular data (Table QA), it{'}s unclear whether, and to what extent existing Table QA models are robust to task-specific perturbations, e.g., replacing key question entities or shuffling table columns. To systematically study the robustness of Table QA models, we propose a benchmark called RobuT, which builds upon existing Table QA datasets (WTQ, WikiSQL-Weak, and SQA) and includes human-annotated adversarial perturbations in terms of table header, table content, and question. Our results indicate that both state-of-the-art Table QA models and large language models (e.g., GPT-3) with few-shot learning falter in these adversarial sets. We propose to address this problem by using large language models to generate adversarial examples to enhance training, which significantly improves the robustness of Table QA models. | [
"Zhao, Yilun",
"Zhao, Chen",
"Nan, Linyong",
"Qi, Zhenting",
"Zhang, Wenlin",
"Tang, Xiangru",
"Mi, Boyu",
"Radev, Dragomir"
] | RobuT: A Systematic Study of Table QA Robustness Against Human-Annotated Adversarial Perturbations | acl-long.334 | Poster | 2306.14321 | [
"https://github.com/yilunzhao/robut"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.335.bib | https://aclanthology.org/2023.acl-long.335/ | @inproceedings{kodner-etal-2023-morphological,
title = "Morphological Inflection: A Reality Check",
author = "Kodner, Jordan and
Payne, Sarah and
Khalifa, Salam and
Liu, Zoey",
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.335",
doi = "10.18653/v1/2023.acl-long.335",
pages = "6082--6101",
abstract = "Morphological inflection is a popular task in sub-word NLP with both practical and cognitive applications. For years now, state-of-the-art systems have reported high, but also highly variable, performance across data sets and languages. We investigate the causes of this high performance and high variability; we find several aspects of data set creation and evaluation which systematically inflate performance and obfuscate differences between languages. To improve generalizability and reliability of results, we propose new data sampling and evaluation strategies that better reflect likely use-cases. Using these new strategies, we make new observations on the generalization abilities of current inflection systems.",
}
| Morphological inflection is a popular task in sub-word NLP with both practical and cognitive applications. For years now, state-of-the-art systems have reported high, but also highly variable, performance across data sets and languages. We investigate the causes of this high performance and high variability; we find several aspects of data set creation and evaluation which systematically inflate performance and obfuscate differences between languages. To improve generalizability and reliability of results, we propose new data sampling and evaluation strategies that better reflect likely use-cases. Using these new strategies, we make new observations on the generalization abilities of current inflection systems. | [
"Kodner, Jordan",
"Payne, Sarah",
"Khalifa, Salam",
"Liu, Zoey"
] | Morphological Inflection: A Reality Check | acl-long.335 | Poster | 2305.15637 | [
"https://github.com/jkodner05/acl2023_realitycheck"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.336.bib | https://aclanthology.org/2023.acl-long.336/ | @inproceedings{ren-etal-2023-tome,
title = "{TOME}: A Two-stage Approach for Model-based Retrieval",
author = "Ren, Ruiyang and
Zhao, Wayne Xin and
Liu, Jing and
Wu, Hua and
Wen, Ji-Rong and
Wang, Haifeng",
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.336",
doi = "10.18653/v1/2023.acl-long.336",
pages = "6102--6114",
abstract = "Recently, model-based retrieval has emerged as a new paradigm in text retrieval that discards the index in the traditional retrieval model and instead memorizes the candidate corpora using model parameters. This design employs a sequence-to-sequence paradigm to generate document identifiers, which enables the complete capture of the relevance between queries and documents and simplifies the classic index-retrieval-rerank pipeline. Despite its attractive qualities, there remain several major challenges in model-based retrieval, including the discrepancy between pre-training and fine-tuning, and the discrepancy between training and inference. To deal with the above challenges, we propose a novel two-stage model-based retrieval approach called TOME, which makes two major technical contributions, including the utilization of tokenized URLs as identifiers and the design of a two-stage generation architecture. We also propose a number of training strategies to deal with the training difficulty as the corpus size increases. Extensive experiments and analysis on MS MARCO and Natural Questions demonstrate the effectiveness of our proposed approach, and we investigate the scaling laws of TOME by examining various influencing factors.",
}
| Recently, model-based retrieval has emerged as a new paradigm in text retrieval that discards the index in the traditional retrieval model and instead memorizes the candidate corpora using model parameters. This design employs a sequence-to-sequence paradigm to generate document identifiers, which enables the complete capture of the relevance between queries and documents and simplifies the classic index-retrieval-rerank pipeline. Despite its attractive qualities, there remain several major challenges in model-based retrieval, including the discrepancy between pre-training and fine-tuning, and the discrepancy between training and inference. To deal with the above challenges, we propose a novel two-stage model-based retrieval approach called TOME, which makes two major technical contributions, including the utilization of tokenized URLs as identifiers and the design of a two-stage generation architecture. We also propose a number of training strategies to deal with the training difficulty as the corpus size increases. Extensive experiments and analysis on MS MARCO and Natural Questions demonstrate the effectiveness of our proposed approach, and we investigate the scaling laws of TOME by examining various influencing factors. | [
"Ren, Ruiyang",
"Zhao, Wayne Xin",
"Liu, Jing",
"Wu, Hua",
"Wen, Ji-Rong",
"Wang, Haifeng"
] | TOME: A Two-stage Approach for Model-based Retrieval | acl-long.336 | Poster | 2305.11161 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.337.bib | https://aclanthology.org/2023.acl-long.337/ | @inproceedings{palma-gomez-etal-2023-using,
title = "Using Neural Machine Translation for Generating Diverse Challenging Exercises for Language Learner",
author = "Palma Gomez, Frank and
Panda, Subhadarshi and
Flor, Michael and
Rozovskaya, Alla",
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.337",
doi = "10.18653/v1/2023.acl-long.337",
pages = "6115--6129",
abstract = "We propose a novel approach to automatically generate distractors for cloze exercises for English language learners, using round-trip neural machine translation. A carrier sentence is translated from English into another (pivot) language and back, and distractors are produced by aligning the original sentence with its round-trip translation. We make use of 16 linguistically-diverse pivots and generate hundreds of translation hypotheses in each direction. We show that using hundreds of translations allows us to generate a rich set of challenging distractors. Moreover, we find that typologically unrelated language pivots contribute more diverse candidate distractors, compared to language pivots that are closely related. We further evaluate the use of machine translation systems of varying quality and find that better quality MT systems produce more challenging distractors. Finally, we conduct a study with language learners, demonstrating that the automatically generated distractors are of the same difficulty as the gold distractors produced by human experts.",
}
| We propose a novel approach to automatically generate distractors for cloze exercises for English language learners, using round-trip neural machine translation. A carrier sentence is translated from English into another (pivot) language and back, and distractors are produced by aligning the original sentence with its round-trip translation. We make use of 16 linguistically-diverse pivots and generate hundreds of translation hypotheses in each direction. We show that using hundreds of translations allows us to generate a rich set of challenging distractors. Moreover, we find that typologically unrelated language pivots contribute more diverse candidate distractors, compared to language pivots that are closely related. We further evaluate the use of machine translation systems of varying quality and find that better quality MT systems produce more challenging distractors. Finally, we conduct a study with language learners, demonstrating that the automatically generated distractors are of the same difficulty as the gold distractors produced by human experts. | [
"Palma Gomez, Frank",
"P",
"a, Subhadarshi",
"Flor, Michael",
"Rozovskaya, Alla"
] | Using Neural Machine Translation for Generating Diverse Challenging Exercises for Language Learner | acl-long.337 | Oral | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.338.bib | https://aclanthology.org/2023.acl-long.338/ | @inproceedings{plenz-etal-2023-similarity,
title = "Similarity-weighted Construction of Contextualized Commonsense Knowledge Graphs for Knowledge-intense Argumentation Tasks",
author = "Plenz, Moritz and
Opitz, Juri and
Heinisch, Philipp and
Cimiano, Philipp and
Frank, Anette",
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.338",
doi = "10.18653/v1/2023.acl-long.338",
pages = "6130--6158",
abstract = "Arguments often do not make explicit how a conclusion follows from its premises. To compensate for this lack, we enrich arguments with structured background knowledge to support knowledge-intense argumentation tasks. We present a new unsupervised method for constructing Contextualized Commonsense Knowledge Graphs (CCKGs) that selects contextually relevant knowledge from large knowledge graphs (KGs) efficiently and at high quality. Our work goes beyond context-insensitive knowledge extraction heuristics by computing semantic similarity between KG triplets and textual arguments. Using these triplet similarities as weights, we extract contextualized knowledge paths that connect a conclusion to its premise, while maximizing similarity to the argument. We combine multiple paths into a CCKG that we optionally prune to reduce noise and raise precision. Intrinsic evaluation of the quality of our graphs shows that our method is effective for (re)constructing human explanation graphs. Manual evaluations in a large-scale knowledge selection setup verify high recall and precision of implicit CSK in the CCKGs. Finally, we demonstrate the effectiveness of CCKGs in a knowledge-insensitive argument quality rating task, outperforming strong baselines and rivaling a GPT-3 based system.",
}
| Arguments often do not make explicit how a conclusion follows from its premises. To compensate for this lack, we enrich arguments with structured background knowledge to support knowledge-intense argumentation tasks. We present a new unsupervised method for constructing Contextualized Commonsense Knowledge Graphs (CCKGs) that selects contextually relevant knowledge from large knowledge graphs (KGs) efficiently and at high quality. Our work goes beyond context-insensitive knowledge extraction heuristics by computing semantic similarity between KG triplets and textual arguments. Using these triplet similarities as weights, we extract contextualized knowledge paths that connect a conclusion to its premise, while maximizing similarity to the argument. We combine multiple paths into a CCKG that we optionally prune to reduce noise and raise precision. Intrinsic evaluation of the quality of our graphs shows that our method is effective for (re)constructing human explanation graphs. Manual evaluations in a large-scale knowledge selection setup verify high recall and precision of implicit CSK in the CCKGs. Finally, we demonstrate the effectiveness of CCKGs in a knowledge-insensitive argument quality rating task, outperforming strong baselines and rivaling a GPT-3 based system. | [
"Plenz, Moritz",
"Opitz, Juri",
"Heinisch, Philipp",
"Cimiano, Philipp",
"Frank, Anette"
] | Similarity-weighted Construction of Contextualized Commonsense Knowledge Graphs for Knowledge-intense Argumentation Tasks | acl-long.338 | Poster | 2305.08495 | [
"https://github.com/heidelberg-nlp/cckg"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.339.bib | https://aclanthology.org/2023.acl-long.339/ | @inproceedings{klein-nabi-2023-micse,
title = "mi{CSE}: Mutual Information Contrastive Learning for Low-shot Sentence Embeddings",
author = "Klein, Tassilo and
Nabi, Moin",
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.339",
doi = "10.18653/v1/2023.acl-long.339",
pages = "6159--6177",
abstract = "This paper presents miCSE, a mutual information-based contrastive learning framework that significantly advances the state-of-the-art in few-shot sentence embedding. The proposed approach imposes alignment between the attention pattern of different views during contrastive learning. Learning sentence embeddings with miCSE entails enforcing the structural consistency across augmented views for every sentence, making contrastive self-supervised learning more sample efficient. As a result, the proposed approach shows strong performance in the few-shot learning domain. While it achieves superior results compared to state-of-the-art methods on multiple benchmarks in few-shot learning, it is comparable in the full-shot scenario. This study opens up avenues for efficient self-supervised learning methods that are more robust than current contrastive methods for sentence embedding.",
}
| This paper presents miCSE, a mutual information-based contrastive learning framework that significantly advances the state-of-the-art in few-shot sentence embedding. The proposed approach imposes alignment between the attention pattern of different views during contrastive learning. Learning sentence embeddings with miCSE entails enforcing the structural consistency across augmented views for every sentence, making contrastive self-supervised learning more sample efficient. As a result, the proposed approach shows strong performance in the few-shot learning domain. While it achieves superior results compared to state-of-the-art methods on multiple benchmarks in few-shot learning, it is comparable in the full-shot scenario. This study opens up avenues for efficient self-supervised learning methods that are more robust than current contrastive methods for sentence embedding. | [
"Klein, Tassilo",
"Nabi, Moin"
] | miCSE: Mutual Information Contrastive Learning for Low-shot Sentence Embeddings | acl-long.339 | Poster | 2211.04928 | [
"https://github.com/sap-samples/acl2023-micse"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.340.bib | https://aclanthology.org/2023.acl-long.340/ | @inproceedings{sharma-etal-2023-learning,
title = "Learning Non-linguistic Skills without Sacrificing Linguistic Proficiency",
author = "Sharma, Mandar and
Muralidhar, Nikhil and
Ramakrishnan, Naren",
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.340",
doi = "10.18653/v1/2023.acl-long.340",
pages = "6178--6191",
abstract = "The field of Math-NLP has witnessed significant growth in recent years, motivated by the desire to expand LLM performance to the leaning of non-linguistic notions (numerals, and subsequently, arithmetic reasoning). However, non-linguistic skill injection typically comes at a cost for LLMs: it leads to catastrophic forgetting of core linguistic skills, a consequence that often remains unaddressed in the literature. As Math-NLP has been able to create LLMs that can closely approximate the mathematical skills of a grade schooler or the arithmetic reasoning skills of a calculator, the practicality of these models fail if they concomitantly shed their linguistic capabilities. In this work, we take a closer look into the phenomena of catastrophic forgetting as it pertains to LLMs and subsequently offer a novel framework for non-linguistic skill injection for LLMs based on information-theoretic interventions and skill-specific losses that enable the learning of strict arithmetic reasoning. Our model outperforms the state-of-the-art both on injected non-linguistic skills and on linguistic knowledge retention, and does so with a fraction of the non-linguistic training data (1/4) and zero additional synthetic linguistic training data.",
}
| The field of Math-NLP has witnessed significant growth in recent years, motivated by the desire to expand LLM performance to the leaning of non-linguistic notions (numerals, and subsequently, arithmetic reasoning). However, non-linguistic skill injection typically comes at a cost for LLMs: it leads to catastrophic forgetting of core linguistic skills, a consequence that often remains unaddressed in the literature. As Math-NLP has been able to create LLMs that can closely approximate the mathematical skills of a grade schooler or the arithmetic reasoning skills of a calculator, the practicality of these models fail if they concomitantly shed their linguistic capabilities. In this work, we take a closer look into the phenomena of catastrophic forgetting as it pertains to LLMs and subsequently offer a novel framework for non-linguistic skill injection for LLMs based on information-theoretic interventions and skill-specific losses that enable the learning of strict arithmetic reasoning. Our model outperforms the state-of-the-art both on injected non-linguistic skills and on linguistic knowledge retention, and does so with a fraction of the non-linguistic training data (1/4) and zero additional synthetic linguistic training data. | [
"Sharma, M",
"ar",
"Muralidhar, Nikhil",
"Ramakrishnan, Naren"
] | Learning Non-linguistic Skills without Sacrificing Linguistic Proficiency | acl-long.340 | Poster | 2305.08246 | [
"https://github.com/mandar-sharma/skill-lm"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.341.bib | https://aclanthology.org/2023.acl-long.341/ | @inproceedings{singh-etal-2023-forgotten,
title = "Forgotten Knowledge: Examining the Citational Amnesia in {NLP}",
author = "Singh, Janvijay and
Rungta, Mukund and
Yang, Diyi and
Mohammad, Saif",
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.341",
doi = "10.18653/v1/2023.acl-long.341",
pages = "6192--6208",
abstract = "Citing papers is the primary method through which modern scientific writing discusses and builds on past work. Collectively, citing a diverse set of papers (in time and area of study) is an indicator of how widely the community is reading. Yet, there is little work looking at broad temporal patterns of citation. This work systematically and empirically examines: How far back in time do we tend to go to cite papers? How has that changed over time, and what factors correlate with this citational attention/amnesia? We chose NLP as our domain of interest and analyzed approximately 71.5K papers to show and quantify several key trends in citation. Notably, around 62{\%} of cited papers are from the immediate five years prior to publication, whereas only about 17{\%} are more than ten years old. Furthermore, we show that the median age and age diversity of cited papers were steadily increasing from 1990 to 2014, but since then, the trend has reversed, and current NLP papers have an all-time low temporal citation diversity. Finally, we show that unlike the 1990s, the highly cited papers in the last decade were also papers with the least citation diversity, likely contributing to the intense (and arguably harmful) recency focus. Code, data, and a demo are available on the project homepage.",
}
| Citing papers is the primary method through which modern scientific writing discusses and builds on past work. Collectively, citing a diverse set of papers (in time and area of study) is an indicator of how widely the community is reading. Yet, there is little work looking at broad temporal patterns of citation. This work systematically and empirically examines: How far back in time do we tend to go to cite papers? How has that changed over time, and what factors correlate with this citational attention/amnesia? We chose NLP as our domain of interest and analyzed approximately 71.5K papers to show and quantify several key trends in citation. Notably, around 62{\%} of cited papers are from the immediate five years prior to publication, whereas only about 17{\%} are more than ten years old. Furthermore, we show that the median age and age diversity of cited papers were steadily increasing from 1990 to 2014, but since then, the trend has reversed, and current NLP papers have an all-time low temporal citation diversity. Finally, we show that unlike the 1990s, the highly cited papers in the last decade were also papers with the least citation diversity, likely contributing to the intense (and arguably harmful) recency focus. Code, data, and a demo are available on the project homepage. | [
"Singh, Janvijay",
"Rungta, Mukund",
"Yang, Diyi",
"Mohammad, Saif"
] | Forgotten Knowledge: Examining the Citational Amnesia in NLP | acl-long.341 | Poster | 2305.18554 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.342.bib | https://aclanthology.org/2023.acl-long.342/ | @inproceedings{du-nguyen-2023-measuring,
title = "Measuring the Instability of Fine-Tuning",
author = "Du, Yupei and
Nguyen, 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.342",
doi = "10.18653/v1/2023.acl-long.342",
pages = "6209--6230",
abstract = "Fine-tuning pre-trained language models on downstream tasks with varying random seeds has been shown to be unstable, especially on small datasets. Many previous studies have investigated this instability and proposed methods to mitigate it. However, most of these studies only used the standard deviation of performance scores (SD) as their measure, which is a narrow characterization of instability. In this paper, we analyze SD and six other measures quantifying instability of different granularity levels. Moreover, we propose a systematic evaluation framework of these measures{'} validity. Finally, we analyze the consistency and difference between different measures by reassessing existing instability mitigation methods. We hope our results will inform better measurements of the fine-tuning instability.",
}
| Fine-tuning pre-trained language models on downstream tasks with varying random seeds has been shown to be unstable, especially on small datasets. Many previous studies have investigated this instability and proposed methods to mitigate it. However, most of these studies only used the standard deviation of performance scores (SD) as their measure, which is a narrow characterization of instability. In this paper, we analyze SD and six other measures quantifying instability of different granularity levels. Moreover, we propose a systematic evaluation framework of these measures{'} validity. Finally, we analyze the consistency and difference between different measures by reassessing existing instability mitigation methods. We hope our results will inform better measurements of the fine-tuning instability. | [
"Du, Yupei",
"Nguyen, Dong"
] | Measuring the Instability of Fine-Tuning | acl-long.342 | Poster | 2302.07778 | [
"https://github.com/nlpsoc/instability_measurement"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.343.bib | https://aclanthology.org/2023.acl-long.343/ | @inproceedings{fleisig-etal-2023-fairprism,
title = "{F}air{P}rism: Evaluating Fairness-Related Harms in Text Generation",
author = "Fleisig, Eve and
Amstutz, Aubrie and
Atalla, Chad and
Blodgett, Su Lin and
Daum{\'e} III, Hal and
Olteanu, Alexandra and
Sheng, Emily and
Vann, Dan and
Wallach, Hanna",
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.343",
doi = "10.18653/v1/2023.acl-long.343",
pages = "6231--6251",
abstract = "It is critical to measure and mitigate fairness-related harms caused by AI text generation systems, including stereotyping and demeaning harms. To that end, we introduce FairPrism, a dataset of 5,000 examples of AI-generated English text with detailed human annotations covering a diverse set of harms relating to gender and sexuality. FairPrism aims to address several limitations of existing datasets for measuring and mitigating fairness-related harms, including improved transparency, clearer specification of dataset coverage, and accounting for annotator disagreement and harms that are context-dependent. FairPrism{'}s annotations include the extent of stereotyping and demeaning harms, the demographic groups targeted, and appropriateness for different applications. The annotations also include specific harms that occur in interactive contexts and harms that raise normative concerns when the {``}speaker{''} is an AI system. Due to its precision and granularity, FairPrism can be used to diagnose (1) the types of fairness-related harms that AI text generation systems cause, and (2) the potential limitations of mitigation methods, both of which we illustrate through case studies. Finally, the process we followed to develop FairPrism offers a recipe for building improved datasets for measuring and mitigating harms caused by AI systems.",
}
| It is critical to measure and mitigate fairness-related harms caused by AI text generation systems, including stereotyping and demeaning harms. To that end, we introduce FairPrism, a dataset of 5,000 examples of AI-generated English text with detailed human annotations covering a diverse set of harms relating to gender and sexuality. FairPrism aims to address several limitations of existing datasets for measuring and mitigating fairness-related harms, including improved transparency, clearer specification of dataset coverage, and accounting for annotator disagreement and harms that are context-dependent. FairPrism{'}s annotations include the extent of stereotyping and demeaning harms, the demographic groups targeted, and appropriateness for different applications. The annotations also include specific harms that occur in interactive contexts and harms that raise normative concerns when the {``}speaker{''} is an AI system. Due to its precision and granularity, FairPrism can be used to diagnose (1) the types of fairness-related harms that AI text generation systems cause, and (2) the potential limitations of mitigation methods, both of which we illustrate through case studies. Finally, the process we followed to develop FairPrism offers a recipe for building improved datasets for measuring and mitigating harms caused by AI systems. | [
"Fleisig, Eve",
"Amstutz, Aubrie",
"Atalla, Chad",
"Blodgett, Su Lin",
"Daum{\\'e} III, Hal",
"Olteanu, Alex",
"ra",
"Sheng, Emily",
"Vann, Dan",
"Wallach, Hanna"
] | FairPrism: Evaluating Fairness-Related Harms in Text Generation | acl-long.343 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.344.bib | https://aclanthology.org/2023.acl-long.344/ | @inproceedings{roit-etal-2023-factually,
title = "Factually Consistent Summarization via Reinforcement Learning with Textual Entailment Feedback",
author = "Roit, Paul and
Ferret, Johan and
Shani, Lior and
Aharoni, Roee and
Cideron, Geoffrey and
Dadashi, Robert and
Geist, Matthieu and
Girgin, Sertan and
Hussenot, Leonard and
Keller, Orgad and
Momchev, Nikola and
Ramos Garea, Sabela and
Stanczyk, Piotr and
Vieillard, Nino and
Bachem, Olivier and
Elidan, Gal and
Hassidim, Avinatan and
Pietquin, Olivier and
Szpektor, Idan",
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.344",
doi = "10.18653/v1/2023.acl-long.344",
pages = "6252--6272",
abstract = "Despite the seeming success of contemporary grounded text generation systems, they often tend to generate factually inconsistent text with respect to their input. This phenomenon is emphasized in tasks like summarization, in which the generated summaries should be corroborated by their source article. In this work we leverage recent progress on textual entailment models to directly address this problem for abstractive summarization systems. We use reinforcement learning with reference-free, textual-entailment rewards to optimize for factual consistency and explore the ensuing trade-offs, as improved consistency may come at the cost of less informative or more extractive summaries. Our results, according to both automatic metrics and human evaluation, show that our method considerably improves the faithfulness, salience and conciseness of the generated summaries.",
}
| Despite the seeming success of contemporary grounded text generation systems, they often tend to generate factually inconsistent text with respect to their input. This phenomenon is emphasized in tasks like summarization, in which the generated summaries should be corroborated by their source article. In this work we leverage recent progress on textual entailment models to directly address this problem for abstractive summarization systems. We use reinforcement learning with reference-free, textual-entailment rewards to optimize for factual consistency and explore the ensuing trade-offs, as improved consistency may come at the cost of less informative or more extractive summaries. Our results, according to both automatic metrics and human evaluation, show that our method considerably improves the faithfulness, salience and conciseness of the generated summaries. | [
"Roit, Paul",
"Ferret, Johan",
"Shani, Lior",
"Aharoni, Roee",
"Cideron, Geoffrey",
"Dadashi, Robert",
"Geist, Matthieu",
"Girgin, Sertan",
"Hussenot, Leonard",
"Keller, Orgad",
"Momchev, Nikola",
"Ramos Garea, Sabela",
"Stanczyk, Piotr",
"Vieillard, Nino",
"Bachem, Olivier",
"Elidan, Gal",
"Hassidim, Avinatan",
"Pietquin, Olivier",
"Szpektor, Idan"
] | Factually Consistent Summarization via Reinforcement Learning with Textual Entailment Feedback | acl-long.344 | Poster | 2306.00186 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.345.bib | https://aclanthology.org/2023.acl-long.345/ | @inproceedings{wu-etal-2023-simmc,
title = "{SIMMC}-{VR}: A Task-oriented Multimodal Dialog Dataset with Situated and Immersive {VR} Streams",
author = "Wu, Te-Lin and
Kottur, Satwik and
Madotto, Andrea and
Azab, Mahmoud and
Rodriguez, Pedro and
Damavandi, Babak and
Peng, Nanyun and
Moon, Seungwhan",
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.345",
doi = "10.18653/v1/2023.acl-long.345",
pages = "6273--6291",
abstract = "Building an AI assistant that can seamlessly converse and instruct humans, in a user-centric situated scenario, requires several essential abilities:(1) spatial and temporal understanding of the situated and real-time user scenes,(2) capability of grounding the actively perceived visuals of users to conversation contexts,and (3) conversational reasoning over past utterances to perform just-in-time assistance. However, we currently lack a large-scale benchmark that captures user{--}assistant interactions with all of the aforementioned features. To this end, we propose SIMMC-VR, an extension of the SIMMC-2.0 dataset, to a video-grounded task-oriented dialog dataset that captures real-world AI-assisted user scenarios in VR.We propose a novel data collection paradigm that involves(1) generating object-centric multimodal dialog flows with egocentric visual streams and visually-grounded templates,and (2) manually paraphrasing the simulated dialogs for naturalness and diversity while preserving multimodal dependencies. To measure meaningful progress in the field, we propose four tasks to address the new challenges in SIMMC-VR, which require complex spatial-temporal dialog reasoning in active egocentric scenes. We benchmark the proposed tasks with strong multimodal models, and highlight the key capabilities that current models lack for future research directions.",
}
| Building an AI assistant that can seamlessly converse and instruct humans, in a user-centric situated scenario, requires several essential abilities:(1) spatial and temporal understanding of the situated and real-time user scenes,(2) capability of grounding the actively perceived visuals of users to conversation contexts,and (3) conversational reasoning over past utterances to perform just-in-time assistance. However, we currently lack a large-scale benchmark that captures user{--}assistant interactions with all of the aforementioned features. To this end, we propose SIMMC-VR, an extension of the SIMMC-2.0 dataset, to a video-grounded task-oriented dialog dataset that captures real-world AI-assisted user scenarios in VR.We propose a novel data collection paradigm that involves(1) generating object-centric multimodal dialog flows with egocentric visual streams and visually-grounded templates,and (2) manually paraphrasing the simulated dialogs for naturalness and diversity while preserving multimodal dependencies. To measure meaningful progress in the field, we propose four tasks to address the new challenges in SIMMC-VR, which require complex spatial-temporal dialog reasoning in active egocentric scenes. We benchmark the proposed tasks with strong multimodal models, and highlight the key capabilities that current models lack for future research directions. | [
"Wu, Te-Lin",
"Kottur, Satwik",
"Madotto, Andrea",
"Azab, Mahmoud",
"Rodriguez, Pedro",
"Damav",
"i, Babak",
"Peng, Nanyun",
"Moon, Seungwhan"
] | SIMMC-VR: A Task-oriented Multimodal Dialog Dataset with Situated and Immersive VR Streams | acl-long.345 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.346.bib | https://aclanthology.org/2023.acl-long.346/ | @inproceedings{tanwar-etal-2023-multilingual,
title = "Multilingual {LLM}s are Better Cross-lingual In-context Learners with Alignment",
author = "Tanwar, Eshaan and
Dutta, Subhabrata and
Borthakur, Manish and
Chakraborty, Tanmoy",
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.346",
doi = "10.18653/v1/2023.acl-long.346",
pages = "6292--6307",
abstract = "In-context learning (ICL) unfolds as large language models become capable of inferring test labels conditioned on a few labeled samples without any gradient update. ICL-enabled large language models provide a promising step forward toward bypassing recurrent annotation costs in a low-resource setting. Yet, only a handful of past studies have explored ICL in a cross-lingual setting, in which the need for transferring label-knowledge from a high-resource language to a low-resource one is immensely crucial. To bridge the gap, we provide the first in-depth analysis of ICL for cross-lingual text classification. We find that the prevalent mode of selecting random input-label pairs to construct the prompt-context is severely limited in the case of cross-lingual ICL, primarily due to the lack of alignment in the input as well as the output spaces. To mitigate this, we propose a novel prompt construction strategy {---} Cross-lingual In-context Source Target Alignment (X-InSTA). With an injected coherence in the semantics of the input examples and a task-based alignment across the source and target languages, X-InSTA is able to outperform random prompt selection by a large margin across three different tasks using 44 different cross-lingual pairs.",
}
| In-context learning (ICL) unfolds as large language models become capable of inferring test labels conditioned on a few labeled samples without any gradient update. ICL-enabled large language models provide a promising step forward toward bypassing recurrent annotation costs in a low-resource setting. Yet, only a handful of past studies have explored ICL in a cross-lingual setting, in which the need for transferring label-knowledge from a high-resource language to a low-resource one is immensely crucial. To bridge the gap, we provide the first in-depth analysis of ICL for cross-lingual text classification. We find that the prevalent mode of selecting random input-label pairs to construct the prompt-context is severely limited in the case of cross-lingual ICL, primarily due to the lack of alignment in the input as well as the output spaces. To mitigate this, we propose a novel prompt construction strategy {---} Cross-lingual In-context Source Target Alignment (X-InSTA). With an injected coherence in the semantics of the input examples and a task-based alignment across the source and target languages, X-InSTA is able to outperform random prompt selection by a large margin across three different tasks using 44 different cross-lingual pairs. | [
"Tanwar, Eshaan",
"Dutta, Subhabrata",
"Borthakur, Manish",
"Chakraborty, Tanmoy"
] | Multilingual LLMs are Better Cross-lingual In-context Learners with Alignment | acl-long.346 | Oral | 2305.05940 | [
"https://github.com/eshaant/x-insta"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.347.bib | https://aclanthology.org/2023.acl-long.347/ | @inproceedings{sanyal-etal-2023-apollo,
title = "{APOLLO}: A Simple Approach for Adaptive Pretraining of Language Models for Logical Reasoning",
author = "Sanyal, Soumya and
Xu, Yichong and
Wang, Shuohang and
Yang, Ziyi and
Pryzant, Reid and
Yu, Wenhao and
Zhu, Chenguang and
Ren, Xiang",
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.347",
doi = "10.18653/v1/2023.acl-long.347",
pages = "6308--6321",
abstract = "Logical reasoning over text is an important ability that requires understanding the semantics of the text and reasoning through them to arrive at correct inferences. Prior works on pretraining language models to improve the logical reasoning ability require complex processing of training data (e.g., aligning symbolic knowledge to text), yielding task-specific data augmentation that is not easy to adapt to any general text corpus. In this work, we propose APOLLO, a simple adaptive pretraining approach to improve the logical reasoning skills of language models. We select a subset of Wikipedia for adaptive pretraining using a set of logical inference keywords as filter words. Further, we propose two self-supervised loss functions for training. First, we modify the masked language modeling loss only to mask specific parts-of-speech words that likely require higher-order reasoning to predict them. Second, we propose a sentence-level classification loss that teaches the model to distinguish between entailment and contradiction types of sentences. The proposed pretraining paradigm is both simple and independent of task formats. We demonstrate the effectiveness of APOLLO by comparing it with prior baselines on two logical reasoning datasets. APOLLO performs comparably on ReClor and outperforms baselines on LogiQA.",
}
| Logical reasoning over text is an important ability that requires understanding the semantics of the text and reasoning through them to arrive at correct inferences. Prior works on pretraining language models to improve the logical reasoning ability require complex processing of training data (e.g., aligning symbolic knowledge to text), yielding task-specific data augmentation that is not easy to adapt to any general text corpus. In this work, we propose APOLLO, a simple adaptive pretraining approach to improve the logical reasoning skills of language models. We select a subset of Wikipedia for adaptive pretraining using a set of logical inference keywords as filter words. Further, we propose two self-supervised loss functions for training. First, we modify the masked language modeling loss only to mask specific parts-of-speech words that likely require higher-order reasoning to predict them. Second, we propose a sentence-level classification loss that teaches the model to distinguish between entailment and contradiction types of sentences. The proposed pretraining paradigm is both simple and independent of task formats. We demonstrate the effectiveness of APOLLO by comparing it with prior baselines on two logical reasoning datasets. APOLLO performs comparably on ReClor and outperforms baselines on LogiQA. | [
"Sanyal, Soumya",
"Xu, Yichong",
"Wang, Shuohang",
"Yang, Ziyi",
"Pryzant, Reid",
"Yu, Wenhao",
"Zhu, Chenguang",
"Ren, Xiang"
] | APOLLO: A Simple Approach for Adaptive Pretraining of Language Models for Logical Reasoning | acl-long.347 | Poster | 2212.09282 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.348.bib | https://aclanthology.org/2023.acl-long.348/ | @inproceedings{pal-etal-2023-multitabqa,
title = "{M}ulti{T}ab{QA}: Generating Tabular Answers for Multi-Table Question Answering",
author = "Pal, Vaishali and
Yates, Andrew and
Kanoulas, Evangelos and
de Rijke, Maarten",
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.348",
doi = "10.18653/v1/2023.acl-long.348",
pages = "6322--6334",
abstract = "Recent advances in tabular question answering (QA) with large language models are constrained in their coverage and only answer questions over a single table. However, real-world queries are complex in nature, often over multiple tables in a relational database or web page. Single table questions do not involve common table operations such as set operations, Cartesian products (joins), or nested queries. Furthermore, multi-table operations often result in a tabular output, which necessitates table generation capabilities of tabular QA models. To fill this gap, we propose a new task of answering questions over multiple tables. Our model, MultiTabQA, not only answers questions over multiple tables, but also generalizes to generate tabular answers. To enable effective training, we build a pre-training dataset comprising of 132,645 SQL queries and tabular answers. Further, we evaluate the generated tables by introducing table-specific metrics of varying strictness assessing various levels of granularity of the table structure. MultiTabQA outperforms state-of-the-art single table QA models adapted to a multi-table QA setting by finetuning on three datasets: Spider, Atis and GeoQuery.",
}
| Recent advances in tabular question answering (QA) with large language models are constrained in their coverage and only answer questions over a single table. However, real-world queries are complex in nature, often over multiple tables in a relational database or web page. Single table questions do not involve common table operations such as set operations, Cartesian products (joins), or nested queries. Furthermore, multi-table operations often result in a tabular output, which necessitates table generation capabilities of tabular QA models. To fill this gap, we propose a new task of answering questions over multiple tables. Our model, MultiTabQA, not only answers questions over multiple tables, but also generalizes to generate tabular answers. To enable effective training, we build a pre-training dataset comprising of 132,645 SQL queries and tabular answers. Further, we evaluate the generated tables by introducing table-specific metrics of varying strictness assessing various levels of granularity of the table structure. MultiTabQA outperforms state-of-the-art single table QA models adapted to a multi-table QA setting by finetuning on three datasets: Spider, Atis and GeoQuery. | [
"Pal, Vaishali",
"Yates, Andrew",
"Kanoulas, Evangelos",
"de Rijke, Maarten"
] | MultiTabQA: Generating Tabular Answers for Multi-Table Question Answering | acl-long.348 | Poster | 2305.12820 | [
"https://github.com/kolk/multitabqa"
] | https://huggingface.co/papers/2305.12820 | 0 | 0 | 0 | 4 | 1 | [
"vaishali/multitabqa-base",
"vaishali/multitabqa-base-sql",
"vaishali/multitabqa-base-atis",
"vaishali/multitabqa-base-geoquery"
] | [] | [] |
https://aclanthology.org/2023.acl-long.349.bib | https://aclanthology.org/2023.acl-long.349/ | @inproceedings{li-etal-2023-copy,
title = "To Copy Rather Than Memorize: A Vertical Learning Paradigm for Knowledge Graph Completion",
author = "Li, Rui and
Chen, Xu and
Li, Chaozhuo and
Shen, Yanming and
Zhao, Jianan and
Wang, Yujing and
Han, Weihao and
Sun, Hao and
Deng, Weiwei and
Zhang, Qi and
Xie, Xing",
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.349",
doi = "10.18653/v1/2023.acl-long.349",
pages = "6335--6347",
abstract = "Embedding models have shown great power in knowledge graph completion (KGC) task. By learning structural constraints for each training triple, these methods implicitly memorize intrinsic relation rules to infer missing links. However, this paper points out that the multi-hop relation rules are hard to be reliably memorized due to the inherent deficiencies of such implicit memorization strategy, making embedding models underperform in predicting links between distant entity pairs. To alleviate this problem, we present Vertical Learning Paradigm (VLP), which extends embedding models by allowing to explicitly copy target information from related factual triples for more accurate prediction. Rather than solely relying on the implicit memory, VLP directly provides additional cues to improve the generalization ability of embedding models, especially making the distant link prediction significantly easier. Moreover, we also propose a novel relative distance based negative sampling technique (ReD) for more effective optimization. Experiments demonstrate the validity and generality of our proposals on two standard benchmarks. Our code is available at \url{https://github.com/rui9812/VLP}.",
}
| Embedding models have shown great power in knowledge graph completion (KGC) task. By learning structural constraints for each training triple, these methods implicitly memorize intrinsic relation rules to infer missing links. However, this paper points out that the multi-hop relation rules are hard to be reliably memorized due to the inherent deficiencies of such implicit memorization strategy, making embedding models underperform in predicting links between distant entity pairs. To alleviate this problem, we present Vertical Learning Paradigm (VLP), which extends embedding models by allowing to explicitly copy target information from related factual triples for more accurate prediction. Rather than solely relying on the implicit memory, VLP directly provides additional cues to improve the generalization ability of embedding models, especially making the distant link prediction significantly easier. Moreover, we also propose a novel relative distance based negative sampling technique (ReD) for more effective optimization. Experiments demonstrate the validity and generality of our proposals on two standard benchmarks. Our code is available at \url{https://github.com/rui9812/VLP}. | [
"Li, Rui",
"Chen, Xu",
"Li, Chaozhuo",
"Shen, Yanming",
"Zhao, Jianan",
"Wang, Yujing",
"Han, Weihao",
"Sun, Hao",
"Deng, Weiwei",
"Zhang, Qi",
"Xie, Xing"
] | To Copy Rather Than Memorize: A Vertical Learning Paradigm for Knowledge Graph Completion | acl-long.349 | Poster | 2305.14126 | [
"https://github.com/rui9812/vlp"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.350.bib | https://aclanthology.org/2023.acl-long.350/ | @inproceedings{wang-etal-2023-coad,
title = "{C}o{AD}: Automatic Diagnosis through Symptom and Disease Collaborative Generation",
author = "Wang, Huimin and
Kwan, Wai Chung and
Wong, Kam-Fai and
Zheng, Yefeng",
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.350",
doi = "10.18653/v1/2023.acl-long.350",
pages = "6348--6361",
abstract = "Automatic diagnosis (AD), a critical application of AI in healthcare, employs machine learning techniques to assist doctors in gathering patient symptom information for precise disease diagnosis. The Transformer-based method utilizes an input symptom sequence, predicts itself through auto-regression, and employs the hidden state of the final symptom to determine the disease. Despite its simplicity and superior performance demonstrated, a decline in disease diagnosis accuracy is observed caused by 1) a mismatch between symptoms observed during training and generation, and 2) the effect of different symptom orders on disease prediction. To address the above obstacles, we introduce the CoAD, a novel disease and symptom collaborative generation framework, which incorporates several key innovations to improve AD: 1) aligning sentence-level disease labels with multiple possible symptom inquiry steps to bridge the gap between training and generation; 2) expanding symptom labels for each sub-sequence of symptoms to enhance annotation and eliminate the effect of symptom order; 3) developing a repeated symptom input schema to effectively and efficiently learn the expanded disease and symptom labels. We evaluate the CoAD framework using four datasets, including three public and one private, and demonstrate that it achieves an average 2.3{\%} improvement over previous state-of-the-art results in automatic disease diagnosis. For reproducibility, we release the code and data at \url{https://github.com/KwanWaiChung/coad}.",
}
| Automatic diagnosis (AD), a critical application of AI in healthcare, employs machine learning techniques to assist doctors in gathering patient symptom information for precise disease diagnosis. The Transformer-based method utilizes an input symptom sequence, predicts itself through auto-regression, and employs the hidden state of the final symptom to determine the disease. Despite its simplicity and superior performance demonstrated, a decline in disease diagnosis accuracy is observed caused by 1) a mismatch between symptoms observed during training and generation, and 2) the effect of different symptom orders on disease prediction. To address the above obstacles, we introduce the CoAD, a novel disease and symptom collaborative generation framework, which incorporates several key innovations to improve AD: 1) aligning sentence-level disease labels with multiple possible symptom inquiry steps to bridge the gap between training and generation; 2) expanding symptom labels for each sub-sequence of symptoms to enhance annotation and eliminate the effect of symptom order; 3) developing a repeated symptom input schema to effectively and efficiently learn the expanded disease and symptom labels. We evaluate the CoAD framework using four datasets, including three public and one private, and demonstrate that it achieves an average 2.3{\%} improvement over previous state-of-the-art results in automatic disease diagnosis. For reproducibility, we release the code and data at \url{https://github.com/KwanWaiChung/coad}. | [
"Wang, Huimin",
"Kwan, Wai Chung",
"Wong, Kam-Fai",
"Zheng, Yefeng"
] | CoAD: Automatic Diagnosis through Symptom and Disease Collaborative Generation | acl-long.350 | Oral | 2307.08290 | [
"https://github.com/kwanwaichung/coad"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.351.bib | https://aclanthology.org/2023.acl-long.351/ | @inproceedings{dai-etal-2023-long,
title = "Long-Tailed Question Answering in an Open World",
author = "Dai, Yi and
Lang, Hao and
Zheng, Yinhe and
Huang, Fei and
Li, Yongbin",
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.351",
doi = "10.18653/v1/2023.acl-long.351",
pages = "6362--6382",
abstract = "Real-world data often have an open long-tailed distribution, and building a unified QA model supporting various tasks is vital for practical QA applications. However, it is non-trivial to extend previous QA approaches since they either require access to seen tasks of adequate samples or do not explicitly model samples from unseen tasks. In this paper, we define Open Long-Tailed QA (OLTQA) as learning from long-tailed distributed data and optimizing performance over seen and unseen QA tasks. We propose an OLTQA model that encourages knowledge sharing between head, tail and unseen tasks, and explicitly mines knowledge from a large pre-trained language model (LM).Specifically, we organize our model through a pool of fine-grained components and dynamically combine these components for an input to facilitate knowledge sharing.A retrieve-then-rerank frame is further introduced to select in-context examples, which guild the LM to generate text that express knowledge for QA tasks. Moreover, a two-stage training approach is introduced to pre-train the framework by knowledge distillation (KD) from the LM and then jointly train the frame and a QA model through an adaptive mutual KD method. On a large-scale OLTQA dataset we curate from 43 existing QA datasets, our model consistently outperforms the state-of-the-art.",
}
| Real-world data often have an open long-tailed distribution, and building a unified QA model supporting various tasks is vital for practical QA applications. However, it is non-trivial to extend previous QA approaches since they either require access to seen tasks of adequate samples or do not explicitly model samples from unseen tasks. In this paper, we define Open Long-Tailed QA (OLTQA) as learning from long-tailed distributed data and optimizing performance over seen and unseen QA tasks. We propose an OLTQA model that encourages knowledge sharing between head, tail and unseen tasks, and explicitly mines knowledge from a large pre-trained language model (LM).Specifically, we organize our model through a pool of fine-grained components and dynamically combine these components for an input to facilitate knowledge sharing.A retrieve-then-rerank frame is further introduced to select in-context examples, which guild the LM to generate text that express knowledge for QA tasks. Moreover, a two-stage training approach is introduced to pre-train the framework by knowledge distillation (KD) from the LM and then jointly train the frame and a QA model through an adaptive mutual KD method. On a large-scale OLTQA dataset we curate from 43 existing QA datasets, our model consistently outperforms the state-of-the-art. | [
"Dai, Yi",
"Lang, Hao",
"Zheng, Yinhe",
"Huang, Fei",
"Li, Yongbin"
] | Long-Tailed Question Answering in an Open World | acl-long.351 | Poster | 2305.06557 | [
"https://github.com/alibabaresearch/damo-convai"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.352.bib | https://aclanthology.org/2023.acl-long.352/ | @inproceedings{ratner-etal-2023-parallel,
title = "Parallel Context Windows for Large Language Models",
author = "Ratner, Nir and
Levine, Yoav and
Belinkov, Yonatan and
Ram, Ori and
Magar, Inbal and
Abend, Omri and
Karpas, Ehud and
Shashua, Amnon and
Leyton-Brown, Kevin and
Shoham, Yoav",
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.352",
doi = "10.18653/v1/2023.acl-long.352",
pages = "6383--6402",
abstract = "When applied to processing long text, Large Language Models (LLMs) are limited by their context window. Existing efforts to address this limitation involve training specialized architectures, and cannot be easily applied to off- the-shelf LLMs. We present Parallel Context Windows (PCW), a method that alleviates the context window restriction for any off-the-shelf LLM without further training. The key to the approach is to carve a long context into chunks ({``}windows{''}), restrict the attention mechanism to apply only within each window, and re-use the positional embeddings across the windows. Our main results test the PCW approach on in-context learning with models that range in size between 750 million and 178 billion parameters, and show substantial improvements for tasks with diverse input and output spaces. We show additional benefits in other settings where long context windows may be beneficial: multi-hop questions and retrieval-augmented question answering with multiple retrieved documents. Our results highlight Parallel Context Windows as a promising method for applying off-the-shelf LLMs in a range of settings that require long text sequences. We make our code publicly available at \url{https://github.com/ai21labs/parallel-context-windows}.",
}
| When applied to processing long text, Large Language Models (LLMs) are limited by their context window. Existing efforts to address this limitation involve training specialized architectures, and cannot be easily applied to off- the-shelf LLMs. We present Parallel Context Windows (PCW), a method that alleviates the context window restriction for any off-the-shelf LLM without further training. The key to the approach is to carve a long context into chunks ({``}windows{''}), restrict the attention mechanism to apply only within each window, and re-use the positional embeddings across the windows. Our main results test the PCW approach on in-context learning with models that range in size between 750 million and 178 billion parameters, and show substantial improvements for tasks with diverse input and output spaces. We show additional benefits in other settings where long context windows may be beneficial: multi-hop questions and retrieval-augmented question answering with multiple retrieved documents. Our results highlight Parallel Context Windows as a promising method for applying off-the-shelf LLMs in a range of settings that require long text sequences. We make our code publicly available at \url{https://github.com/ai21labs/parallel-context-windows}. | [
"Ratner, Nir",
"Levine, Yoav",
"Belinkov, Yonatan",
"Ram, Ori",
"Magar, Inbal",
"Abend, Omri",
"Karpas, Ehud",
"Shashua, Amnon",
"Leyton-Brown, Kevin",
"Shoham, Yoav"
] | Parallel Context Windows for Large Language Models | acl-long.352 | Poster | 2212.10947 | [
"https://github.com/AI21Labs/Parallel-Context-Windows"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.353.bib | https://aclanthology.org/2023.acl-long.353/ | @inproceedings{nawrot-etal-2023-efficient,
title = "Efficient Transformers with Dynamic Token Pooling",
author = "Nawrot, Piotr and
Chorowski, Jan and
Lancucki, Adrian and
Ponti, Edoardo Maria",
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.353",
doi = "10.18653/v1/2023.acl-long.353",
pages = "6403--6417",
abstract = "Transformers achieve unrivalled performance in modelling language, but remain inefficient in terms of memory and time complexity. A possible remedy is to reduce the sequence length in the intermediate layers by pooling fixed-length segments of tokens. Nevertheless, natural units of meaning, such as words or phrases, display varying sizes. To address this mismatch, we equip language models with a dynamic-pooling mechanism, which predicts segment boundaries in an autoregressive fashion. We compare several methods to infer boundaries, including end-to-end learning through stochastic re-parameterisation, supervised learning (based on segmentations from subword tokenizers or spikes in conditional entropy), as well as linguistically motivated boundaries. We perform character-level evaluation on texts from multiple datasets and morphologically diverse languages. The results demonstrate that dynamic pooling, which jointly segments and models language, is both faster and more accurate than vanilla Transformers and fixed-length pooling within the same computational budget.",
}
| Transformers achieve unrivalled performance in modelling language, but remain inefficient in terms of memory and time complexity. A possible remedy is to reduce the sequence length in the intermediate layers by pooling fixed-length segments of tokens. Nevertheless, natural units of meaning, such as words or phrases, display varying sizes. To address this mismatch, we equip language models with a dynamic-pooling mechanism, which predicts segment boundaries in an autoregressive fashion. We compare several methods to infer boundaries, including end-to-end learning through stochastic re-parameterisation, supervised learning (based on segmentations from subword tokenizers or spikes in conditional entropy), as well as linguistically motivated boundaries. We perform character-level evaluation on texts from multiple datasets and morphologically diverse languages. The results demonstrate that dynamic pooling, which jointly segments and models language, is both faster and more accurate than vanilla Transformers and fixed-length pooling within the same computational budget. | [
"Nawrot, Piotr",
"Chorowski, Jan",
"Lancucki, Adrian",
"Ponti, Edoardo Maria"
] | Efficient Transformers with Dynamic Token Pooling | acl-long.353 | Poster | 2211.09761 | [
"https://github.com/piotrnawrot/dynamic-pooling"
] | https://huggingface.co/papers/2211.09761 | 0 | 0 | 1 | 4 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.354.bib | https://aclanthology.org/2023.acl-long.354/ | @inproceedings{chen-etal-2023-models,
title = "Did the Models Understand Documents? Benchmarking Models for Language Understanding in Document-Level Relation Extraction",
author = "Chen, Haotian and
Chen, Bingsheng and
Zhou, Xiangdong",
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.354",
doi = "10.18653/v1/2023.acl-long.354",
pages = "6418--6435",
abstract = "Document-level relation extraction (DocRE) attracts more research interest recently. While models achieve consistent performance gains in DocRE, their underlying decision rules are still understudied: Do they make the right predictions according to rationales? In this paper, we take the first step toward answering this question and then introduce a new perspective on comprehensively evaluating a model. Specifically, we first conduct annotations to provide the rationales considered by humans in DocRE. Then, we conduct investigations and discover the fact that: In contrast to humans, the representative state-of-the-art (SOTA) models in DocRE exhibit different reasoning processes. Through our proposed RE-specific attacks, we next demonstrate that the significant discrepancy in decision rules between models and humans severely damages the robustness of models. After that, we introduce mean average precision (MAP) to evaluate the understanding and reasoning capabilities of models. According to the extensive experimental results, we finally appeal to future work to consider evaluating the understanding ability of models because the improved ability renders models more trustworthy and robust to be deployed in real-world scenarios. We make our annotations and code publicly available.",
}
| Document-level relation extraction (DocRE) attracts more research interest recently. While models achieve consistent performance gains in DocRE, their underlying decision rules are still understudied: Do they make the right predictions according to rationales? In this paper, we take the first step toward answering this question and then introduce a new perspective on comprehensively evaluating a model. Specifically, we first conduct annotations to provide the rationales considered by humans in DocRE. Then, we conduct investigations and discover the fact that: In contrast to humans, the representative state-of-the-art (SOTA) models in DocRE exhibit different reasoning processes. Through our proposed RE-specific attacks, we next demonstrate that the significant discrepancy in decision rules between models and humans severely damages the robustness of models. After that, we introduce mean average precision (MAP) to evaluate the understanding and reasoning capabilities of models. According to the extensive experimental results, we finally appeal to future work to consider evaluating the understanding ability of models because the improved ability renders models more trustworthy and robust to be deployed in real-world scenarios. We make our annotations and code publicly available. | [
"Chen, Haotian",
"Chen, Bingsheng",
"Zhou, Xiangdong"
] | Did the Models Understand Documents? Benchmarking Models for Language Understanding in Document-Level Relation Extraction | acl-long.354 | Poster | 2306.11386 | [
"https://github.com/hytn/docred-hwe"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.355.bib | https://aclanthology.org/2023.acl-long.355/ | @inproceedings{jain-etal-2023-contraclm,
title = "{C}ontra{CLM}: Contrastive Learning For Causal Language Model",
author = "Jain, Nihal and
Zhang, Dejiao and
Ahmad, Wasi Uddin and
Wang, Zijian and
Nan, Feng and
Li, Xiaopeng and
Tan, Ming and
Nallapati, Ramesh and
Ray, Baishakhi and
Bhatia, Parminder and
Ma, Xiaofei and
Xiang, 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.355",
doi = "10.18653/v1/2023.acl-long.355",
pages = "6436--6459",
abstract = "Despite exciting progress in causal language models, the expressiveness of their representations is largely limited due to poor discrimination ability. To remedy this issue, we present CONTRACLM, a novel contrastive learning framework at both the token-level and the sequence-level. We assess CONTRACLM on a variety of downstream tasks. We show that CONTRACLM enhances the discrimination of representations and bridges the gap with encoder-only models, which makes causal language models better suited for tasks beyond language generation. Specifically, we attain 44{\%} relative improvement on the Semantic Textual Similarity tasks and 34{\%} on Code-to-Code Search tasks. Furthermore, by improving the expressiveness of representations, CONTRACLM also boosts the source code generation capability with 9{\%} relative improvement on execution accuracy on the HumanEval benchmark.",
}
| Despite exciting progress in causal language models, the expressiveness of their representations is largely limited due to poor discrimination ability. To remedy this issue, we present CONTRACLM, a novel contrastive learning framework at both the token-level and the sequence-level. We assess CONTRACLM on a variety of downstream tasks. We show that CONTRACLM enhances the discrimination of representations and bridges the gap with encoder-only models, which makes causal language models better suited for tasks beyond language generation. Specifically, we attain 44{\%} relative improvement on the Semantic Textual Similarity tasks and 34{\%} on Code-to-Code Search tasks. Furthermore, by improving the expressiveness of representations, CONTRACLM also boosts the source code generation capability with 9{\%} relative improvement on execution accuracy on the HumanEval benchmark. | [
"Jain, Nihal",
"Zhang, Dejiao",
"Ahmad, Wasi Uddin",
"Wang, Zijian",
"Nan, Feng",
"Li, Xiaopeng",
"Tan, Ming",
"Nallapati, Ramesh",
"Ray, Baishakhi",
"Bhatia, Parminder",
"Ma, Xiaofei",
"Xiang, Bing"
] | ContraCLM: Contrastive Learning For Causal Language Model | acl-long.355 | Poster | 2210.01185 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.356.bib | https://aclanthology.org/2023.acl-long.356/ | @inproceedings{chou-etal-2023-advancing,
title = "Advancing Multi-Criteria {C}hinese Word Segmentation Through Criterion Classification and Denoising",
author = "Chou, Tzu Hsuan and
Lin, Chun-Yi and
Kao, Hung-Yu",
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.356",
doi = "10.18653/v1/2023.acl-long.356",
pages = "6460--6476",
abstract = "Recent research on multi-criteria Chinese word segmentation (MCCWS) mainly focuses on building complex private structures, adding more handcrafted features, or introducing complex optimization processes. In this work, we show that through a simple yet elegant input-hint-based MCCWS model, we can achieve state-of-the-art (SoTA) performances on several datasets simultaneously. We further propose a novel criterion-denoising objective that hurts slightly on F1 score but achieves SoTA recall on out-of-vocabulary words. Our result establishes a simple yet strong baseline for future MCCWS research. Source code is available at \url{https://github.com/IKMLab/MCCWS}.",
}
| Recent research on multi-criteria Chinese word segmentation (MCCWS) mainly focuses on building complex private structures, adding more handcrafted features, or introducing complex optimization processes. In this work, we show that through a simple yet elegant input-hint-based MCCWS model, we can achieve state-of-the-art (SoTA) performances on several datasets simultaneously. We further propose a novel criterion-denoising objective that hurts slightly on F1 score but achieves SoTA recall on out-of-vocabulary words. Our result establishes a simple yet strong baseline for future MCCWS research. Source code is available at \url{https://github.com/IKMLab/MCCWS}. | [
"Chou, Tzu Hsuan",
"Lin, Chun-Yi",
"Kao, Hung-Yu"
] | Advancing Multi-Criteria Chinese Word Segmentation Through Criterion Classification and Denoising | acl-long.356 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.357.bib | https://aclanthology.org/2023.acl-long.357/ | @inproceedings{zhao-etal-2023-infusing,
title = "Infusing Hierarchical Guidance into Prompt Tuning: A Parameter-Efficient Framework for Multi-level Implicit Discourse Relation Recognition",
author = "Zhao, Haodong and
He, Ruifang and
Xiao, Mengnan and
Xu, Jing",
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.357",
doi = "10.18653/v1/2023.acl-long.357",
pages = "6477--6492",
abstract = "Multi-level implicit discourse relation recognition (MIDRR) aims at identifying hierarchical discourse relations among arguments. Previous methods achieve the promotion through fine-tuning PLMs. However, due to the data scarcity and the task gap, the pre-trained feature space cannot be accurately tuned to the task-specific space, which even aggravates the collapse of the vanilla space. Besides, the comprehension of hierarchical semantics for MIDRR makes the conversion much harder. In this paper, we propose a prompt-based Parameter-Efficient Multi-level IDRR (PEMI) framework to solve the above problems. First, we leverage parameter-efficient prompt tuning to drive the inputted arguments to match the pre-trained space and realize the approximation with few parameters. Furthermore, we propose a hierarchical label refining (HLR) method for the prompt verbalizer to deeply integrate hierarchical guidance into the prompt tuning. Finally, our model achieves comparable results on PDTB 2.0 and 3.0 using about 0.1{\%} trainable parameters compared with baselines and the visualization demonstrates the effectiveness of our HLR method.",
}
| Multi-level implicit discourse relation recognition (MIDRR) aims at identifying hierarchical discourse relations among arguments. Previous methods achieve the promotion through fine-tuning PLMs. However, due to the data scarcity and the task gap, the pre-trained feature space cannot be accurately tuned to the task-specific space, which even aggravates the collapse of the vanilla space. Besides, the comprehension of hierarchical semantics for MIDRR makes the conversion much harder. In this paper, we propose a prompt-based Parameter-Efficient Multi-level IDRR (PEMI) framework to solve the above problems. First, we leverage parameter-efficient prompt tuning to drive the inputted arguments to match the pre-trained space and realize the approximation with few parameters. Furthermore, we propose a hierarchical label refining (HLR) method for the prompt verbalizer to deeply integrate hierarchical guidance into the prompt tuning. Finally, our model achieves comparable results on PDTB 2.0 and 3.0 using about 0.1{\%} trainable parameters compared with baselines and the visualization demonstrates the effectiveness of our HLR method. | [
"Zhao, Haodong",
"He, Ruifang",
"Xiao, Mengnan",
"Xu, Jing"
] | Infusing Hierarchical Guidance into Prompt Tuning: A Parameter-Efficient Framework for Multi-level Implicit Discourse Relation Recognition | acl-long.357 | Poster | 2402.15080 | [
"https://github.com/cyclone-joker/idrr_pdtb3_conns"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.358.bib | https://aclanthology.org/2023.acl-long.358/ | @inproceedings{zhan-etal-2023-contrastive,
title = "Contrastive Learning with Adversarial Examples for Alleviating Pathology of Language Model",
author = "Zhan, Pengwei and
Yang, Jing and
Huang, Xiao and
Jing, Chunlei and
Li, Jingying and
Wang, Liming",
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.358",
doi = "10.18653/v1/2023.acl-long.358",
pages = "6493--6508",
abstract = "Neural language models have achieved superior performance. However, these models also suffer from the pathology of overconfidence in the out-of-distribution examples, potentially making the model difficult to interpret and making the interpretation methods fail to provide faithful attributions. In this paper, we explain the model pathology from the view of sentence representation and argue that the counter-intuitive bias degree and direction of the out-of-distribution examples{'} representation cause the pathology. We propose a Contrastive learning regularization method using Adversarial examples for Alleviating the Pathology (ConAAP), which calibrates the sentence representation of out-of-distribution examples. ConAAP generates positive and negative examples following the attribution results and utilizes adversarial examples to introduce direction information in regularization. Experiments show that ConAAP effectively alleviates the model pathology while slightly impacting the generalization ability on in-distribution examples and thus helps interpretation methods obtain more faithful results.",
}
| Neural language models have achieved superior performance. However, these models also suffer from the pathology of overconfidence in the out-of-distribution examples, potentially making the model difficult to interpret and making the interpretation methods fail to provide faithful attributions. In this paper, we explain the model pathology from the view of sentence representation and argue that the counter-intuitive bias degree and direction of the out-of-distribution examples{'} representation cause the pathology. We propose a Contrastive learning regularization method using Adversarial examples for Alleviating the Pathology (ConAAP), which calibrates the sentence representation of out-of-distribution examples. ConAAP generates positive and negative examples following the attribution results and utilizes adversarial examples to introduce direction information in regularization. Experiments show that ConAAP effectively alleviates the model pathology while slightly impacting the generalization ability on in-distribution examples and thus helps interpretation methods obtain more faithful results. | [
"Zhan, Pengwei",
"Yang, Jing",
"Huang, Xiao",
"Jing, Chunlei",
"Li, Jingying",
"Wang, Liming"
] | Contrastive Learning with Adversarial Examples for Alleviating Pathology of Language Model | acl-long.358 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.359.bib | https://aclanthology.org/2023.acl-long.359/ | @inproceedings{toro-isaza-etal-2023-fairy,
title = "Are Fairy Tales Fair? Analyzing Gender Bias in Temporal Narrative Event Chains of Children{'}s Fairy Tales",
author = "Toro Isaza, Paulina and
Xu, Guangxuan and
Oloko, Toye and
Hou, Yufang and
Peng, Nanyun and
Wang, Dakuo",
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.359",
doi = "10.18653/v1/2023.acl-long.359",
pages = "6509--6531",
abstract = "Social biases and stereotypes are embedded in our culture in part through their presence in our stories, as evidenced by the rich history of humanities and social science literature analyzing such biases in children stories. Because these analyses are often conducted manually and at a small scale, such investigations can benefit from the use of more recent natural language processing (NLP) methods that examine social bias in models and data corpora. Our work joins this interdisciplinary effort and makes a unique contribution by taking into account the event narrative structures when analyzing the social bias of stories. We propose a computational pipeline that automatically extracts a story{'}s temporal narrative verb-based event chain for each of its characters as well as character attributes such as gender. We also present a verb-based event annotation scheme that can facilitate bias analysis by including categories such as those that align with traditional stereotypes. Through a case study analyzing gender bias in fairy tales, we demonstrate that our framework can reveal bias in not only the unigram verb-based events in which female and male characters participate but also in the temporal narrative order of such event participation.",
}
| Social biases and stereotypes are embedded in our culture in part through their presence in our stories, as evidenced by the rich history of humanities and social science literature analyzing such biases in children stories. Because these analyses are often conducted manually and at a small scale, such investigations can benefit from the use of more recent natural language processing (NLP) methods that examine social bias in models and data corpora. Our work joins this interdisciplinary effort and makes a unique contribution by taking into account the event narrative structures when analyzing the social bias of stories. We propose a computational pipeline that automatically extracts a story{'}s temporal narrative verb-based event chain for each of its characters as well as character attributes such as gender. We also present a verb-based event annotation scheme that can facilitate bias analysis by including categories such as those that align with traditional stereotypes. Through a case study analyzing gender bias in fairy tales, we demonstrate that our framework can reveal bias in not only the unigram verb-based events in which female and male characters participate but also in the temporal narrative order of such event participation. | [
"Toro Isaza, Paulina",
"Xu, Guangxuan",
"Oloko, Toye",
"Hou, Yufang",
"Peng, Nanyun",
"Wang, Dakuo"
] | Are Fairy Tales Fair? Analyzing Gender Bias in Temporal Narrative Event Chains of Children's Fairy Tales | acl-long.359 | Poster | 2305.16641 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.360.bib | https://aclanthology.org/2023.acl-long.360/ | @inproceedings{zeng-etal-2023-futuretod,
title = "{F}uture{TOD}: Teaching Future Knowledge to Pre-trained Language Model for Task-Oriented Dialogue",
author = "Zeng, Weihao and
He, Keqing and
Wang, Yejie and
Zeng, Chen and
Wang, Jingang and
Xian, Yunsen and
Xu, Weiran",
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.360",
doi = "10.18653/v1/2023.acl-long.360",
pages = "6532--6546",
abstract = "Pre-trained language models based on general text enable huge success in the NLP scenario. But the intrinsical difference of linguistic patterns between general text and task-oriented dialogues makes existing pre-trained language models less useful in practice. Current dialogue pre-training methods rely on a contrastive framework and face the challenges of both selecting true positives and hard negatives. In this paper, we propose a novel dialogue pre-training model, FutureTOD, which distills future knowledge to the representation of the previous dialogue context using a self-training framework. Our intuition is that a good dialogue representation both learns local context information and predicts future information. Extensive experiments on diverse downstream dialogue tasks demonstrate the effectiveness of our model, especially the generalization, robustness, and learning discriminative dialogue representations capabilities.",
}
| Pre-trained language models based on general text enable huge success in the NLP scenario. But the intrinsical difference of linguistic patterns between general text and task-oriented dialogues makes existing pre-trained language models less useful in practice. Current dialogue pre-training methods rely on a contrastive framework and face the challenges of both selecting true positives and hard negatives. In this paper, we propose a novel dialogue pre-training model, FutureTOD, which distills future knowledge to the representation of the previous dialogue context using a self-training framework. Our intuition is that a good dialogue representation both learns local context information and predicts future information. Extensive experiments on diverse downstream dialogue tasks demonstrate the effectiveness of our model, especially the generalization, robustness, and learning discriminative dialogue representations capabilities. | [
"Zeng, Weihao",
"He, Keqing",
"Wang, Yejie",
"Zeng, Chen",
"Wang, Jingang",
"Xian, Yunsen",
"Xu, Weiran"
] | FutureTOD: Teaching Future Knowledge to Pre-trained Language Model for Task-Oriented Dialogue | acl-long.360 | Poster | 2306.10315 | [
"https://github.com/zeng-wh/futuretod"
] | https://huggingface.co/papers/2306.10315 | 1 | 1 | 0 | 7 | 1 | [
"AndrewZeng/futuretod-base-v1.0"
] | [] | [] |
https://aclanthology.org/2023.acl-long.361.bib | https://aclanthology.org/2023.acl-long.361/ | @inproceedings{kazemi-etal-2023-lambada,
title = "{LAMBADA}: Backward Chaining for Automated Reasoning in Natural Language",
author = "Kazemi, Mehran and
Kim, Najoung and
Bhatia, Deepti and
Xu, Xin and
Ramachandran, Deepak",
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.361",
doi = "10.18653/v1/2023.acl-long.361",
pages = "6547--6568",
abstract = "Remarkable progress has been made on automated reasoning with natural text, by using Large Language Models (LLMs) and methods such as Chain-of-Thought prompting and Selection-Inference. These techniques search for proofs in the forward direction from axioms to the conclusion, which suffers from a combinatorial explosion of the search space, and thus high failure rates for problems requiring longer chains of reasoning. The classical automated reasoning literature has shown that reasoning in the backward direction (i.e. from intended conclusion to supporting axioms) is significantly more efficient at proof-finding. Importing this intuition into the LM setting, we develop a Backward Chaining algorithm, called LAMBADA, that decomposes reasoning into four sub-modules, that are simply implemented by few-shot prompted LLM inference. We show that LAMBADA achieves sizable accuracy boosts over state-of-the-art forward reasoning methods on two challenging logical reasoning datasets, particularly when deep and accurate proof chains are required.",
}
| Remarkable progress has been made on automated reasoning with natural text, by using Large Language Models (LLMs) and methods such as Chain-of-Thought prompting and Selection-Inference. These techniques search for proofs in the forward direction from axioms to the conclusion, which suffers from a combinatorial explosion of the search space, and thus high failure rates for problems requiring longer chains of reasoning. The classical automated reasoning literature has shown that reasoning in the backward direction (i.e. from intended conclusion to supporting axioms) is significantly more efficient at proof-finding. Importing this intuition into the LM setting, we develop a Backward Chaining algorithm, called LAMBADA, that decomposes reasoning into four sub-modules, that are simply implemented by few-shot prompted LLM inference. We show that LAMBADA achieves sizable accuracy boosts over state-of-the-art forward reasoning methods on two challenging logical reasoning datasets, particularly when deep and accurate proof chains are required. | [
"Kazemi, Mehran",
"Kim, Najoung",
"Bhatia, Deepti",
"Xu, Xin",
"Ramach",
"ran, Deepak"
] | LAMBADA: Backward Chaining for Automated Reasoning in Natural Language | acl-long.361 | Poster | 2212.13894 | [
""
] | https://huggingface.co/papers/2212.13894 | 1 | 0 | 0 | 5 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.362.bib | https://aclanthology.org/2023.acl-long.362/ | @inproceedings{gao-etal-2023-peacok,
title = "{P}ea{C}o{K}: Persona Commonsense Knowledge for Consistent and Engaging Narratives",
author = "Gao, Silin and
Borges, Beatriz and
Oh, Soyoung and
Bayazit, Deniz and
Kanno, Saya and
Wakaki, Hiromi and
Mitsufuji, Yuki and
Bosselut, Antoine",
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.362",
doi = "10.18653/v1/2023.acl-long.362",
pages = "6569--6591",
abstract = "Sustaining coherent and engaging narratives requires dialogue or storytelling agents to understandhow the personas of speakers or listeners ground the narrative. Specifically, these agents must infer personas of their listeners to produce statements that cater to their interests. They must also learn to maintain consistent speaker personas for themselves throughout the narrative, so that their counterparts feel involved in a realistic conversation or story. However, personas are diverse and complex: they entail large quantities of rich interconnected world knowledge that is challenging to robustly represent in general narrative systems (e.g., a singer is good at singing, and may have attended conservatoire). In this work, we construct a new large-scale persona commonsense knowledge graph, PeaCoK, containing {\textasciitilde}100K human-validated persona facts. Our knowledge graph schematizes five dimensions of persona knowledge identified in previous studies of human interactive behaviours, and distils facts in this schema from both existing commonsense knowledge graphs and large-scale pretrained language models. Our analysis indicates that PeaCoK contains rich and precise world persona inferences that help downstream systems generate more consistent and engaging narratives.",
}
| Sustaining coherent and engaging narratives requires dialogue or storytelling agents to understandhow the personas of speakers or listeners ground the narrative. Specifically, these agents must infer personas of their listeners to produce statements that cater to their interests. They must also learn to maintain consistent speaker personas for themselves throughout the narrative, so that their counterparts feel involved in a realistic conversation or story. However, personas are diverse and complex: they entail large quantities of rich interconnected world knowledge that is challenging to robustly represent in general narrative systems (e.g., a singer is good at singing, and may have attended conservatoire). In this work, we construct a new large-scale persona commonsense knowledge graph, PeaCoK, containing {\textasciitilde}100K human-validated persona facts. Our knowledge graph schematizes five dimensions of persona knowledge identified in previous studies of human interactive behaviours, and distils facts in this schema from both existing commonsense knowledge graphs and large-scale pretrained language models. Our analysis indicates that PeaCoK contains rich and precise world persona inferences that help downstream systems generate more consistent and engaging narratives. | [
"Gao, Silin",
"Borges, Beatriz",
"Oh, Soyoung",
"Bayazit, Deniz",
"Kanno, Saya",
"Wakaki, Hiromi",
"Mitsufuji, Yuki",
"Bosselut, Antoine"
] | PeaCoK: Persona Commonsense Knowledge for Consistent and Engaging Narratives | acl-long.362 | Poster | 2305.02364 | [
"https://github.com/silin159/peacok"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.363.bib | https://aclanthology.org/2023.acl-long.363/ | @inproceedings{cheng-etal-2023-opensr,
title = "{O}pen{SR}: Open-Modality Speech Recognition via Maintaining Multi-Modality Alignment",
author = "Cheng, Xize and
Jin, Tao and
Li, Linjun and
Lin, Wang and
Duan, Xinyu and
Zhao, Zhou",
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.363",
doi = "10.18653/v1/2023.acl-long.363",
pages = "6592--6607",
abstract = "Speech Recognition builds a bridge between the multimedia streaming (audio-only, visual-only or audio-visual) and the corresponding text transcription. However, when training the specific model of new domain, it often gets stuck in the lack of new-domain utterances, especially the labeled visual utterances. To break through this restriction, we attempt to achieve zero-shot modality transfer by maintaining the multi-modality alignment in phoneme space learned with unlabeled multimedia utterances in the high resource domain during the pre-training, and propose a training system Open-modality Speech Recognition (\textbf{OpenSR}) that enables the models trained on a single modality (e.g., audio-only) applicable to more modalities (e.g., visual-only and audio-visual). Furthermore, we employ a cluster-based prompt tuning strategy to handle the domain shift for the scenarios with only common words in the new domain utterances. We demonstrate that OpenSR enables modality transfer from one to any in three different settings (zero-, few- and full-shot), and achieves highly competitive zero-shot performance compared to the existing few-shot and full-shot lip-reading methods. To the best of our knowledge, OpenSR achieves the state-of-the-art performance of word error rate in LRS2 on audio-visual speech recognition and lip-reading with 2.7{\%} and 25.0{\%}, respectively.",
}
| Speech Recognition builds a bridge between the multimedia streaming (audio-only, visual-only or audio-visual) and the corresponding text transcription. However, when training the specific model of new domain, it often gets stuck in the lack of new-domain utterances, especially the labeled visual utterances. To break through this restriction, we attempt to achieve zero-shot modality transfer by maintaining the multi-modality alignment in phoneme space learned with unlabeled multimedia utterances in the high resource domain during the pre-training, and propose a training system Open-modality Speech Recognition (\textbf{OpenSR}) that enables the models trained on a single modality (e.g., audio-only) applicable to more modalities (e.g., visual-only and audio-visual). Furthermore, we employ a cluster-based prompt tuning strategy to handle the domain shift for the scenarios with only common words in the new domain utterances. We demonstrate that OpenSR enables modality transfer from one to any in three different settings (zero-, few- and full-shot), and achieves highly competitive zero-shot performance compared to the existing few-shot and full-shot lip-reading methods. To the best of our knowledge, OpenSR achieves the state-of-the-art performance of word error rate in LRS2 on audio-visual speech recognition and lip-reading with 2.7{\%} and 25.0{\%}, respectively. | [
"Cheng, Xize",
"Jin, Tao",
"Li, Linjun",
"Lin, Wang",
"Duan, Xinyu",
"Zhao, Zhou"
] | OpenSR: Open-Modality Speech Recognition via Maintaining Multi-Modality Alignment | acl-long.363 | Oral | 2306.06410 | [
"https://github.com/exgc/opensr"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.364.bib | https://aclanthology.org/2023.acl-long.364/ | @inproceedings{wang-etal-2023-retrieval,
title = "Retrieval-free Knowledge Injection through Multi-Document Traversal for Dialogue Models",
author = "Wang, Rui and
Bao, Jianzhu and
Mi, Fei and
Chen, Yi and
Wang, Hongru and
Wang, Yasheng and
Li, Yitong and
Shang, Lifeng and
Wong, Kam-Fai and
Xu, Ruifeng",
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.364",
doi = "10.18653/v1/2023.acl-long.364",
pages = "6608--6619",
abstract = "Dialogue models are often enriched with extensive external knowledge to provide informative responses through a retrieval-augmented pipeline. Nevertheless, retrieval-augmented approaches rely on finely annotated retrieval training data and knowledge-grounded response generation data, making it costly to transfer. To tackle this challenge, this paper proposed a retrieval-free approach, KiDG, by automatically turning knowledge documents into simulated multi-turn dialogues through a Multi-Document Traversal algorithm. The simulated knowledge-intensive dialogues constructed by KiDG in one domain can be easily used to train and enhance pre-trained dialogue models{'} knowledge w.r.t. this domain without costly annotation. We conduct extensive experiments comparing retrieval-augmented models and a variety of retrieval-free models. We found that dialogue models enhanced with data simulated with KiDG largely outperform state-of-the-art retrieval-free methods, and it achieves comparable performance compared to retrieval-augmented methods while being better, and cheaper at domain transfer.",
}
| Dialogue models are often enriched with extensive external knowledge to provide informative responses through a retrieval-augmented pipeline. Nevertheless, retrieval-augmented approaches rely on finely annotated retrieval training data and knowledge-grounded response generation data, making it costly to transfer. To tackle this challenge, this paper proposed a retrieval-free approach, KiDG, by automatically turning knowledge documents into simulated multi-turn dialogues through a Multi-Document Traversal algorithm. The simulated knowledge-intensive dialogues constructed by KiDG in one domain can be easily used to train and enhance pre-trained dialogue models{'} knowledge w.r.t. this domain without costly annotation. We conduct extensive experiments comparing retrieval-augmented models and a variety of retrieval-free models. We found that dialogue models enhanced with data simulated with KiDG largely outperform state-of-the-art retrieval-free methods, and it achieves comparable performance compared to retrieval-augmented methods while being better, and cheaper at domain transfer. | [
"Wang, Rui",
"Bao, Jianzhu",
"Mi, Fei",
"Chen, Yi",
"Wang, Hongru",
"Wang, Yasheng",
"Li, Yitong",
"Shang, Lifeng",
"Wong, Kam-Fai",
"Xu, Ruifeng"
] | Retrieval-free Knowledge Injection through Multi-Document Traversal for Dialogue Models | acl-long.364 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.365.bib | https://aclanthology.org/2023.acl-long.365/ | @inproceedings{xu-etal-2023-berm,
title = "{BERM}: Training the Balanced and Extractable Representation for Matching to Improve Generalization Ability of Dense Retrieval",
author = "Xu, Shicheng and
Pang, Liang and
Shen, Huawei and
Cheng, Xueqi",
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.365",
doi = "10.18653/v1/2023.acl-long.365",
pages = "6620--6635",
abstract = "Dense retrieval has shown promise in the first-stage retrieval process when trained on in-domain labeled datasets. However, previous studies have found that dense retrieval is hard to generalize to unseen domains due to its weak modeling of domain-invariant and interpretable feature (i.e., matching signal between two texts, which is the essence of information retrieval). In this paper, we propose a novel method to improve the generalization of dense retrieval via capturing matching signal called BERM. Fully fine-grained expression and query-oriented saliency are two properties of the matching signal. Thus, in BERM, a single passage is segmented into multiple units and two unit-level requirements are proposed for representation as the constraint in training to obtain the effective matching signal. One is semantic unit balance and the other is essential matching unit extractability. Unit-level view and balanced semantics make representation express the text in a fine-grained manner. Essential matching unit extractability makes passage representation sensitive to the given query to extract the pure matching information from the passage containing complex context. Experiments on BEIR show that our method can be effectively combined with different dense retrieval training methods (vanilla, hard negatives mining and knowledge distillation) to improve its generalization ability without any additional inference overhead and target domain data.",
}
| Dense retrieval has shown promise in the first-stage retrieval process when trained on in-domain labeled datasets. However, previous studies have found that dense retrieval is hard to generalize to unseen domains due to its weak modeling of domain-invariant and interpretable feature (i.e., matching signal between two texts, which is the essence of information retrieval). In this paper, we propose a novel method to improve the generalization of dense retrieval via capturing matching signal called BERM. Fully fine-grained expression and query-oriented saliency are two properties of the matching signal. Thus, in BERM, a single passage is segmented into multiple units and two unit-level requirements are proposed for representation as the constraint in training to obtain the effective matching signal. One is semantic unit balance and the other is essential matching unit extractability. Unit-level view and balanced semantics make representation express the text in a fine-grained manner. Essential matching unit extractability makes passage representation sensitive to the given query to extract the pure matching information from the passage containing complex context. Experiments on BEIR show that our method can be effectively combined with different dense retrieval training methods (vanilla, hard negatives mining and knowledge distillation) to improve its generalization ability without any additional inference overhead and target domain data. | [
"Xu, Shicheng",
"Pang, Liang",
"Shen, Huawei",
"Cheng, Xueqi"
] | BERM: Training the Balanced and Extractable Representation for Matching to Improve Generalization Ability of Dense Retrieval | acl-long.365 | Oral | 2305.11052 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.366.bib | https://aclanthology.org/2023.acl-long.366/ | @inproceedings{li-etal-2023-multiview,
title = "Multiview Identifiers Enhanced Generative Retrieval",
author = "Li, Yongqi and
Yang, Nan and
Wang, Liang and
Wei, Furu and
Li, Wenjie",
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.366",
doi = "10.18653/v1/2023.acl-long.366",
pages = "6636--6648",
abstract = "Instead of simply matching a query to pre-existing passages, generative retrieval generates identifier strings of passages as the retrieval target. At a cost, the identifier must be distinctive enough to represent a passage. Current approaches use either a numeric ID or a text piece (such as a title or substrings) as the identifier. However, these identifiers cannot cover a passage{'}s content well. As such, we are motivated to propose a new type of identifier, synthetic identifiers, that are generated based on the content of a passage and could integrate contextualized information that text pieces lack. Furthermore, we simultaneously consider multiview identifiers, including synthetic identifiers, titles, and substrings. These views of identifiers complement each other and facilitate the holistic ranking of passages from multiple perspectives. We conduct a series of experiments on three public datasets, and the results indicate that our proposed approach performs the best in generative retrieval, demonstrating its effectiveness and robustness.",
}
| Instead of simply matching a query to pre-existing passages, generative retrieval generates identifier strings of passages as the retrieval target. At a cost, the identifier must be distinctive enough to represent a passage. Current approaches use either a numeric ID or a text piece (such as a title or substrings) as the identifier. However, these identifiers cannot cover a passage{'}s content well. As such, we are motivated to propose a new type of identifier, synthetic identifiers, that are generated based on the content of a passage and could integrate contextualized information that text pieces lack. Furthermore, we simultaneously consider multiview identifiers, including synthetic identifiers, titles, and substrings. These views of identifiers complement each other and facilitate the holistic ranking of passages from multiple perspectives. We conduct a series of experiments on three public datasets, and the results indicate that our proposed approach performs the best in generative retrieval, demonstrating its effectiveness and robustness. | [
"Li, Yongqi",
"Yang, Nan",
"Wang, Liang",
"Wei, Furu",
"Li, Wenjie"
] | Multiview Identifiers Enhanced Generative Retrieval | acl-long.366 | Poster | 2305.16675 | [
"https://github.com/liyongqi67/minder"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.367.bib | https://aclanthology.org/2023.acl-long.367/ | @inproceedings{blevins-etal-2023-prompting,
title = "Prompting Language Models for Linguistic Structure",
author = "Blevins, Terra and
Gonen, Hila and
Zettlemoyer, Luke",
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.367",
doi = "10.18653/v1/2023.acl-long.367",
pages = "6649--6663",
abstract = "Although pretrained language models (PLMs) can be prompted to perform a wide range of language tasks, it remains an open question how much this ability comes from generalizable linguistic understanding versus surface-level lexical patterns. To test this, we present a structured prompting approach for linguistic structured prediction tasks, allowing us to perform zero- and few-shot sequence tagging with autoregressive PLMs. We evaluate this approach on part-of-speech tagging, named entity recognition, and sentence chunking, demonstrating strong few-shot performance in all cases. We also find that while PLMs contain significant prior knowledge of task labels due to task leakage into the pretraining corpus, structured prompting can also retrieve linguistic structure with arbitrary labels. These findings indicate that the in-context learning ability and linguistic knowledge of PLMs generalizes beyond memorization of their training data.",
}
| Although pretrained language models (PLMs) can be prompted to perform a wide range of language tasks, it remains an open question how much this ability comes from generalizable linguistic understanding versus surface-level lexical patterns. To test this, we present a structured prompting approach for linguistic structured prediction tasks, allowing us to perform zero- and few-shot sequence tagging with autoregressive PLMs. We evaluate this approach on part-of-speech tagging, named entity recognition, and sentence chunking, demonstrating strong few-shot performance in all cases. We also find that while PLMs contain significant prior knowledge of task labels due to task leakage into the pretraining corpus, structured prompting can also retrieve linguistic structure with arbitrary labels. These findings indicate that the in-context learning ability and linguistic knowledge of PLMs generalizes beyond memorization of their training data. | [
"Blevins, Terra",
"Gonen, Hila",
"Zettlemoyer, Luke"
] | Prompting Language Models for Linguistic Structure | acl-long.367 | Poster | 2211.07830 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.368.bib | https://aclanthology.org/2023.acl-long.368/ | @inproceedings{shah-etal-2023-trillion,
title = "Trillion Dollar Words: A New Financial Dataset, Task {\&} Market Analysis",
author = "Shah, Agam and
Paturi, Suvan and
Chava, Sudheer",
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.368",
doi = "10.18653/v1/2023.acl-long.368",
pages = "6664--6679",
abstract = "Monetary policy pronouncements by Federal Open Market Committee (FOMC) are a major driver of financial market returns. We construct the largest tokenized and annotated dataset of FOMC speeches, meeting minutes, and press conference transcripts in order to understand how monetary policy influences financial markets. In this study, we develop a novel task of hawkish-dovish classification and benchmark various pre-trained language models on the proposed dataset. Using the best-performing model (RoBERTa-large), we construct a measure of monetary policy stance for the FOMC document release days. To evaluate the constructed measure, we study its impact on the treasury market, stock market, and macroeconomic indicators. Our dataset, models, and code are publicly available on Huggingface and GitHub under CC BY-NC 4.0 license.",
}
| Monetary policy pronouncements by Federal Open Market Committee (FOMC) are a major driver of financial market returns. We construct the largest tokenized and annotated dataset of FOMC speeches, meeting minutes, and press conference transcripts in order to understand how monetary policy influences financial markets. In this study, we develop a novel task of hawkish-dovish classification and benchmark various pre-trained language models on the proposed dataset. Using the best-performing model (RoBERTa-large), we construct a measure of monetary policy stance for the FOMC document release days. To evaluate the constructed measure, we study its impact on the treasury market, stock market, and macroeconomic indicators. Our dataset, models, and code are publicly available on Huggingface and GitHub under CC BY-NC 4.0 license. | [
"Shah, Agam",
"Paturi, Suvan",
"Chava, Sudheer"
] | Trillion Dollar Words: A New Financial Dataset, Task & Market Analysis | acl-long.368 | Poster | 2305.07972 | [
"https://github.com/gtfintechlab/fomc-hawkish-dovish"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.369.bib | https://aclanthology.org/2023.acl-long.369/ | @inproceedings{zhao-etal-2023-matching,
title = "{RE}-Matching: A Fine-Grained Semantic Matching Method for Zero-Shot Relation Extraction",
author = "Zhao, Jun and
Zhan, WenYu and
Zhao, Xin and
Zhang, Qi and
Gui, Tao and
Wei, Zhongyu and
Wang, Junzhe and
Peng, Minlong and
Sun, Mingming",
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.369",
doi = "10.18653/v1/2023.acl-long.369",
pages = "6680--6691",
abstract = "Semantic matching is a mainstream paradigm of zero-shot relation extraction, which matches a given input with a corresponding label description. The entities in the input should exactly match their hypernyms in the description, while the irrelevant contexts should be ignored when matching. However, general matching methods lack explicit modeling of the above matching pattern. In this work, we propose a fine-grained semantic matching method tailored for zero-shot relation extraction. Guided by the above matching pattern, we decompose the sentence-level similarity score into the entity matching score and context matching score. Considering that not all contextual words contribute equally to the relation semantics, we design a context distillation module to reduce the negative impact of irrelevant components on context matching. Experimental results show that our method achieves higher matching accuracy and more than 10 times faster inference speed, compared with the state-of-the-art methods.",
}
| Semantic matching is a mainstream paradigm of zero-shot relation extraction, which matches a given input with a corresponding label description. The entities in the input should exactly match their hypernyms in the description, while the irrelevant contexts should be ignored when matching. However, general matching methods lack explicit modeling of the above matching pattern. In this work, we propose a fine-grained semantic matching method tailored for zero-shot relation extraction. Guided by the above matching pattern, we decompose the sentence-level similarity score into the entity matching score and context matching score. Considering that not all contextual words contribute equally to the relation semantics, we design a context distillation module to reduce the negative impact of irrelevant components on context matching. Experimental results show that our method achieves higher matching accuracy and more than 10 times faster inference speed, compared with the state-of-the-art methods. | [
"Zhao, Jun",
"Zhan, WenYu",
"Zhao, Xin",
"Zhang, Qi",
"Gui, Tao",
"Wei, Zhongyu",
"Wang, Junzhe",
"Peng, Minlong",
"Sun, Mingming"
] | RE-Matching: A Fine-Grained Semantic Matching Method for Zero-Shot Relation Extraction | acl-long.369 | Poster | 2306.04954 | [
"https://github.com/zweny/re-matching"
] | https://huggingface.co/papers/2306.04954 | 0 | 0 | 0 | 9 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.370.bib | https://aclanthology.org/2023.acl-long.370/ | @inproceedings{lee-etal-2023-square,
title = "{SQ}u{AR}e: A Large-Scale Dataset of Sensitive Questions and Acceptable Responses Created through Human-Machine Collaboration",
author = "Lee, Hwaran and
Hong, Seokhee and
Park, Joonsuk and
Kim, Takyoung and
Cha, Meeyoung and
Choi, Yejin and
Kim, Byoungpil and
Kim, Gunhee and
Lee, Eun-Ju and
Lim, Yong and
Oh, Alice and
Park, Sangchul and
Ha, Jung-Woo",
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.370",
doi = "10.18653/v1/2023.acl-long.370",
pages = "6692--6712",
abstract = "The potential social harms that large language models pose, such as generating offensive content and reinforcing biases, are steeply rising. Existing works focus on coping with this concern while interacting with ill-intentioned users, such as those who explicitly make hate speech or elicit harmful responses. However, discussions on sensitive issues can become toxic even if the users are well-intentioned. For safer models in such scenarios, we present the Sensitive Questions and Acceptable Response (SQuARe) dataset, a large-scale Korean dataset of 49k sensitive questions with 42k acceptable and 46k non-acceptable responses. The dataset was constructed leveraging HyperCLOVA in a human-in-the-loop manner based on real news headlines. Experiments show that acceptable response generation significantly improves for HyperCLOVA and GPT-3, demonstrating the efficacy of this dataset.",
}
| The potential social harms that large language models pose, such as generating offensive content and reinforcing biases, are steeply rising. Existing works focus on coping with this concern while interacting with ill-intentioned users, such as those who explicitly make hate speech or elicit harmful responses. However, discussions on sensitive issues can become toxic even if the users are well-intentioned. For safer models in such scenarios, we present the Sensitive Questions and Acceptable Response (SQuARe) dataset, a large-scale Korean dataset of 49k sensitive questions with 42k acceptable and 46k non-acceptable responses. The dataset was constructed leveraging HyperCLOVA in a human-in-the-loop manner based on real news headlines. Experiments show that acceptable response generation significantly improves for HyperCLOVA and GPT-3, demonstrating the efficacy of this dataset. | [
"Lee, Hwaran",
"Hong, Seokhee",
"Park, Joonsuk",
"Kim, Takyoung",
"Cha, Meeyoung",
"Choi, Yejin",
"Kim, Byoungpil",
"Kim, Gunhee",
"Lee, Eun-Ju",
"Lim, Yong",
"Oh, Alice",
"Park, Sangchul",
"Ha, Jung-Woo"
] | SQuARe: A Large-Scale Dataset of Sensitive Questions and Acceptable Responses Created through Human-Machine Collaboration | acl-long.370 | Oral | 2305.17696 | [
"https://github.com/naver-ai/korean-safety-benchmarks"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.371.bib | https://aclanthology.org/2023.acl-long.371/ | @inproceedings{yoon-etal-2023-towards,
title = "Towards standardizing {K}orean Grammatical Error Correction: Datasets and Annotation",
author = "Yoon, Soyoung and
Park, Sungjoon and
Kim, Gyuwan and
Cho, Junhee and
Park, Kihyo and
Kim, Gyu Tae and
Seo, Minjoon and
Oh, Alice",
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.371",
doi = "10.18653/v1/2023.acl-long.371",
pages = "6713--6742",
abstract = "Research on Korean grammatical error correction (GEC) is limited, compared to other major languages such as English. We attribute this problematic circumstance to the lack of a carefully designed evaluation benchmark for Korean GEC. In this work, we collect three datasets from different sources (Kor-Lang8, Kor-Native, and Kor-Learner) that covers a wide range of Korean grammatical errors. Considering the nature of Korean grammar, We then define 14 error types for Korean and provide KAGAS (Korean Automatic Grammatical error Annotation System), which can automatically annotate error types from parallel corpora. We use KAGAS on our datasets to make an evaluation benchmark for Korean, and present baseline models trained from our datasets. We show that the model trained with our datasets significantly outperforms the currently used statistical Korean GEC system (Hanspell) on a wider range of error types, demonstrating the diversity and usefulness of the datasets. The implementations and datasets are open-sourced.",
}
| Research on Korean grammatical error correction (GEC) is limited, compared to other major languages such as English. We attribute this problematic circumstance to the lack of a carefully designed evaluation benchmark for Korean GEC. In this work, we collect three datasets from different sources (Kor-Lang8, Kor-Native, and Kor-Learner) that covers a wide range of Korean grammatical errors. Considering the nature of Korean grammar, We then define 14 error types for Korean and provide KAGAS (Korean Automatic Grammatical error Annotation System), which can automatically annotate error types from parallel corpora. We use KAGAS on our datasets to make an evaluation benchmark for Korean, and present baseline models trained from our datasets. We show that the model trained with our datasets significantly outperforms the currently used statistical Korean GEC system (Hanspell) on a wider range of error types, demonstrating the diversity and usefulness of the datasets. The implementations and datasets are open-sourced. | [
"Yoon, Soyoung",
"Park, Sungjoon",
"Kim, Gyuwan",
"Cho, Junhee",
"Park, Kihyo",
"Kim, Gyu Tae",
"Seo, Minjoon",
"Oh, Alice"
] | Towards standardizing Korean Grammatical Error Correction: Datasets and Annotation | acl-long.371 | Poster | 2210.14389 | [
"https://github.com/soyoung97/standard_korean_gec"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.372.bib | https://aclanthology.org/2023.acl-long.372/ | @inproceedings{zhou-etal-2023-flame,
title = "{FL}am{E}: Few-shot Learning from Natural Language Explanations",
author = "Zhou, Yangqiaoyu and
Zhang, Yiming and
Tan, Chenhao",
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.372",
doi = "10.18653/v1/2023.acl-long.372",
pages = "6743--6763",
abstract = "Natural language explanations have the potential to provide rich information that in principle guides model reasoning. Yet, recent work by Lampinen et al. has shown limited utility of natural language explanations in improving classification. To effectively learn from explanations, we present FLamE, a two-stage few-shot learning framework that first generates explanations using GPT-3, and then fine-tunes a smaller model (e.g., RoBERTa) with generated explanations. Our experiments on natural language inference demonstrate effectiveness over strong baselines, increasing accuracy by 17.6{\%} over GPT-3 Babbage and 5.7{\%} over GPT-3 Davinci in e-SNLI.Despite improving classification performance, human evaluation surprisingly reveals that the majority of generated explanations does not adequately justify classification decisions. Additional analyses point to the important role of label-specific cues (e.g., {``}not know{''} for the neutral label) in generated explanations.",
}
| Natural language explanations have the potential to provide rich information that in principle guides model reasoning. Yet, recent work by Lampinen et al. has shown limited utility of natural language explanations in improving classification. To effectively learn from explanations, we present FLamE, a two-stage few-shot learning framework that first generates explanations using GPT-3, and then fine-tunes a smaller model (e.g., RoBERTa) with generated explanations. Our experiments on natural language inference demonstrate effectiveness over strong baselines, increasing accuracy by 17.6{\%} over GPT-3 Babbage and 5.7{\%} over GPT-3 Davinci in e-SNLI.Despite improving classification performance, human evaluation surprisingly reveals that the majority of generated explanations does not adequately justify classification decisions. Additional analyses point to the important role of label-specific cues (e.g., {``}not know{''} for the neutral label) in generated explanations. | [
"Zhou, Yangqiaoyu",
"Zhang, Yiming",
"Tan, Chenhao"
] | FLamE: Few-shot Learning from Natural Language Explanations | acl-long.372 | Oral | 2306.08042 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.373.bib | https://aclanthology.org/2023.acl-long.373/ | @inproceedings{chaudhury-etal-2023-learning,
title = "Learning Symbolic Rules over {A}bstract {M}eaning {R}epresentations for Textual Reinforcement Learning",
author = "Chaudhury, Subhajit and
Swaminathan, Sarathkrishna and
Kimura, Daiki and
Sen, Prithviraj and
Murugesan, Keerthiram and
Uceda-Sosa, Rosario and
Tatsubori, Michiaki and
Fokoue, Achille and
Kapanipathi, Pavan and
Munawar, Asim and
Gray, Alexander",
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.373",
doi = "10.18653/v1/2023.acl-long.373",
pages = "6764--6776",
abstract = "Text-based reinforcement learning agents have predominantly been neural network-based models with embeddings-based representation, learning uninterpretable policies that often do not generalize well to unseen games. On the other hand, neuro-symbolic methods, specifically those that leverage an intermediate formal representation, are gaining significant attention in language understanding tasks. This is because of their advantages ranging from inherent interpretability, the lesser requirement of training data, and being generalizable in scenarios with unseen data. Therefore, in this paper, we propose a modular, NEuro-Symbolic Textual Agent (NESTA) that combines a generic semantic parser with a rule induction system to learn abstract interpretable rules as policies. Our experiments on established text-based game benchmarks show that the proposed NESTA method outperforms deep reinforcement learning-based techniques by achieving better generalization to unseen test games and learning from fewer training interactions.",
}
| Text-based reinforcement learning agents have predominantly been neural network-based models with embeddings-based representation, learning uninterpretable policies that often do not generalize well to unseen games. On the other hand, neuro-symbolic methods, specifically those that leverage an intermediate formal representation, are gaining significant attention in language understanding tasks. This is because of their advantages ranging from inherent interpretability, the lesser requirement of training data, and being generalizable in scenarios with unseen data. Therefore, in this paper, we propose a modular, NEuro-Symbolic Textual Agent (NESTA) that combines a generic semantic parser with a rule induction system to learn abstract interpretable rules as policies. Our experiments on established text-based game benchmarks show that the proposed NESTA method outperforms deep reinforcement learning-based techniques by achieving better generalization to unseen test games and learning from fewer training interactions. | [
"Chaudhury, Subhajit",
"Swaminathan, Sarathkrishna",
"Kimura, Daiki",
"Sen, Prithviraj",
"Murugesan, Keerthiram",
"Uceda-Sosa, Rosario",
"Tatsubori, Michiaki",
"Fokoue, Achille",
"Kapanipathi, Pavan",
"Munawar, Asim",
"Gray, Alex",
"er"
] | Learning Symbolic Rules over Abstract Meaning Representations for Textual Reinforcement Learning | acl-long.373 | Poster | 2307.02689 | [
"https://github.com/ibm/loa"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.374.bib | https://aclanthology.org/2023.acl-long.374/ | @inproceedings{xu-etal-2023-counterfactual,
title = "Counterfactual Debiasing for Fact Verification",
author = "Xu, Weizhi and
Liu, Qiang and
Wu, Shu and
Wang, Liang",
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.374",
doi = "10.18653/v1/2023.acl-long.374",
pages = "6777--6789",
abstract = "Fact verification aims to automatically judge the veracity of a claim according to several pieces of evidence. Due to the manual construction of datasets, spurious correlations between claim patterns and its veracity (i.e., biases) inevitably exist. Recent studies show that models usually learn such biases instead of understanding the semantic relationship between the claim and evidence. Existing debiasing works can be roughly divided into data-augmentation-based and weight-regularization-based pipeline, where the former is inflexible and the latter relies on the uncertain output on the training stage. Unlike previous works, we propose a novel method from a counterfactual view, namely CLEVER, which is augmentation-free and mitigates biases on the inference stage. Specifically, we train a claim-evidence fusion model and a claim-only model independently. Then, we obtain the final prediction via subtracting output of the claim-only model from output of the claim-evidence fusion model, which counteracts biases in two outputs so that the unbiased part is highlighted. Comprehensive experiments on several datasets have demonstrated the effectiveness of CLEVER.",
}
| Fact verification aims to automatically judge the veracity of a claim according to several pieces of evidence. Due to the manual construction of datasets, spurious correlations between claim patterns and its veracity (i.e., biases) inevitably exist. Recent studies show that models usually learn such biases instead of understanding the semantic relationship between the claim and evidence. Existing debiasing works can be roughly divided into data-augmentation-based and weight-regularization-based pipeline, where the former is inflexible and the latter relies on the uncertain output on the training stage. Unlike previous works, we propose a novel method from a counterfactual view, namely CLEVER, which is augmentation-free and mitigates biases on the inference stage. Specifically, we train a claim-evidence fusion model and a claim-only model independently. Then, we obtain the final prediction via subtracting output of the claim-only model from output of the claim-evidence fusion model, which counteracts biases in two outputs so that the unbiased part is highlighted. Comprehensive experiments on several datasets have demonstrated the effectiveness of CLEVER. | [
"Xu, Weizhi",
"Liu, Qiang",
"Wu, Shu",
"Wang, Liang"
] | Counterfactual Debiasing for Fact Verification | acl-long.374 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.375.bib | https://aclanthology.org/2023.acl-long.375/ | @inproceedings{watson-etal-2023-social,
title = "What social attitudes about gender does {BERT} encode? Leveraging insights from psycholinguistics",
author = "Watson, Julia and
Beekhuizen, Barend and
Stevenson, Suzanne",
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.375",
doi = "10.18653/v1/2023.acl-long.375",
pages = "6790--6809",
abstract = "Much research has sought to evaluate the degree to which large language models reflect social biases. We complement such work with an approach to elucidating the connections between language model predictions and people{'}s social attitudes. We show how word preferences in a large language model reflect social attitudes about gender, using two datasets from human experiments that found differences in gendered or gender neutral word choices by participants with differing views on gender (progressive, moderate, or conservative). We find that the language model BERT takes into account factors that shape human lexical choice of such language, but may not weigh those factors in the same way people do. Moreover, we show that BERT{'}s predictions most resemble responses from participants with moderate to conservative views on gender. Such findings illuminate how a language model: (1) may differ from people in how it deploys words that signal gender, and (2) may prioritize some social attitudes over others.",
}
| Much research has sought to evaluate the degree to which large language models reflect social biases. We complement such work with an approach to elucidating the connections between language model predictions and people{'}s social attitudes. We show how word preferences in a large language model reflect social attitudes about gender, using two datasets from human experiments that found differences in gendered or gender neutral word choices by participants with differing views on gender (progressive, moderate, or conservative). We find that the language model BERT takes into account factors that shape human lexical choice of such language, but may not weigh those factors in the same way people do. Moreover, we show that BERT{'}s predictions most resemble responses from participants with moderate to conservative views on gender. Such findings illuminate how a language model: (1) may differ from people in how it deploys words that signal gender, and (2) may prioritize some social attitudes over others. | [
"Watson, Julia",
"Beekhuizen, Barend",
"Stevenson, Suzanne"
] | What social attitudes about gender does BERT encode? Leveraging insights from psycholinguistics | acl-long.375 | Oral | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.376.bib | https://aclanthology.org/2023.acl-long.376/ | @inproceedings{zheng-etal-2023-rethinking,
title = "Rethinking Multimodal Entity and Relation Extraction from a Translation Point of View",
author = "Zheng, Changmeng and
Feng, Junhao and
Cai, Yi and
Wei, Xiaoyong and
Li, Qing",
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.376",
doi = "10.18653/v1/2023.acl-long.376",
pages = "6810--6824",
abstract = "We revisit the multimodal entity and relation extraction from a translation point of view. Special attention is paid on the misalignment issue in text-image datasets which may mislead the learning. We are motivated by the fact that the cross-modal misalignment is a similar problem of cross-lingual divergence issue in machine translation. The problem can then be transformed and existing solutions can be borrowed by treating a text and its paired image as the translation to each other. We implement a multimodal back-translation using diffusion-based generative models for pseudo-paralleled pairs and a divergence estimator by constructing a high-resource corpora as a bridge for low-resource learners. Fine-grained confidence scores are generated to indicate both types and degrees of alignments with which better representations are obtained. The method has been validated in the experiments by outperforming 14 state-of-the-art methods in both entity and relation extraction tasks. The source code is available at \url{https://github.com/thecharm/TMR}.",
}
| We revisit the multimodal entity and relation extraction from a translation point of view. Special attention is paid on the misalignment issue in text-image datasets which may mislead the learning. We are motivated by the fact that the cross-modal misalignment is a similar problem of cross-lingual divergence issue in machine translation. The problem can then be transformed and existing solutions can be borrowed by treating a text and its paired image as the translation to each other. We implement a multimodal back-translation using diffusion-based generative models for pseudo-paralleled pairs and a divergence estimator by constructing a high-resource corpora as a bridge for low-resource learners. Fine-grained confidence scores are generated to indicate both types and degrees of alignments with which better representations are obtained. The method has been validated in the experiments by outperforming 14 state-of-the-art methods in both entity and relation extraction tasks. The source code is available at \url{https://github.com/thecharm/TMR}. | [
"Zheng, Changmeng",
"Feng, Junhao",
"Cai, Yi",
"Wei, Xiaoyong",
"Li, Qing"
] | Rethinking Multimodal Entity and Relation Extraction from a Translation Point of View | acl-long.376 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.377.bib | https://aclanthology.org/2023.acl-long.377/ | @inproceedings{zhu-etal-2023-annotating,
title = "Annotating and Detecting Fine-grained Factual Errors for Dialogue Summarization",
author = "Zhu, Rongxin and
Qi, Jianzhong and
Lau, Jey Han",
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.377",
doi = "10.18653/v1/2023.acl-long.377",
pages = "6825--6845",
abstract = "A series of datasets and models have been proposed for summaries generated for well-formatted documents such as news articles. Dialogue summaries, however, have been under explored. In this paper, we present the first dataset with fine-grained factual error annotations named DIASUMFACT. We define fine-grained factual error detection as a sentence-level multi-label classification problem, and weevaluate two state-of-the-art (SOTA) models on our dataset. Both models yield sub-optimal results, with a macro-averaged F1 score of around 0.25 over 6 error classes. We further propose an unsupervised model ENDERANKER via candidate ranking using pretrained encoder-decoder models. Our model performs on par with the SOTA models while requiring fewer resources. These observations confirm the challenges in detecting factual errors from dialogue summaries, which call for further studies, for which our dataset and results offer a solid foundation.",
}
| A series of datasets and models have been proposed for summaries generated for well-formatted documents such as news articles. Dialogue summaries, however, have been under explored. In this paper, we present the first dataset with fine-grained factual error annotations named DIASUMFACT. We define fine-grained factual error detection as a sentence-level multi-label classification problem, and weevaluate two state-of-the-art (SOTA) models on our dataset. Both models yield sub-optimal results, with a macro-averaged F1 score of around 0.25 over 6 error classes. We further propose an unsupervised model ENDERANKER via candidate ranking using pretrained encoder-decoder models. Our model performs on par with the SOTA models while requiring fewer resources. These observations confirm the challenges in detecting factual errors from dialogue summaries, which call for further studies, for which our dataset and results offer a solid foundation. | [
"Zhu, Rongxin",
"Qi, Jianzhong",
"Lau, Jey Han"
] | Annotating and Detecting Fine-grained Factual Errors for Dialogue Summarization | acl-long.377 | Poster | 2305.16548 | [
"https://github.com/731935354/dia-sum-fact"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.378.bib | https://aclanthology.org/2023.acl-long.378/ | @inproceedings{chen-etal-2023-improving-robustness,
title = "Improving the Robustness of Summarization Systems with Dual Augmentation",
author = "Chen, Xiuying and
Long, Guodong and
Tao, Chongyang and
Li, Mingzhe and
Gao, Xin and
Zhang, Chengqi and
Zhang, Xiangliang",
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.378",
doi = "10.18653/v1/2023.acl-long.378",
pages = "6846--6857",
abstract = "A robust summarization system should be able to capture the gist of the document, regardless of the specific word choices or noise in the input. In this work, we first explore the summarization models{'} robustness against perturbations including word-level synonym substitution and noise. To create semantic-consistent substitutes, we propose a SummAttacker, which is an efficient approach to generating adversarial samples based on pre-trained language models. Experimental results show that state-of-the-art summarization models have a significant decrease in performance on adversarial and noisy test sets. Next, we analyze the vulnerability of the summarization systems and explore improving the robustness by data augmentation. Specifically, the first vulnerability factor we found is the low diversity of the training inputs. Correspondingly, we expose the encoder to more diverse cases created by SummAttacker in the input space. The second factor is the vulnerability of the decoder, and we propose an augmentation in the latent space of the decoder to improve its robustness. Concretely, we create virtual cases by manifold softmixing two decoder hidden states of similar semantic meanings. Experimental results on Gigaword and CNN/DM datasets demonstrate that our approach achieves significant improvements over strong baselines and exhibits higher robustness on noisy, attacked, and clean datasets",
}
| A robust summarization system should be able to capture the gist of the document, regardless of the specific word choices or noise in the input. In this work, we first explore the summarization models{'} robustness against perturbations including word-level synonym substitution and noise. To create semantic-consistent substitutes, we propose a SummAttacker, which is an efficient approach to generating adversarial samples based on pre-trained language models. Experimental results show that state-of-the-art summarization models have a significant decrease in performance on adversarial and noisy test sets. Next, we analyze the vulnerability of the summarization systems and explore improving the robustness by data augmentation. Specifically, the first vulnerability factor we found is the low diversity of the training inputs. Correspondingly, we expose the encoder to more diverse cases created by SummAttacker in the input space. The second factor is the vulnerability of the decoder, and we propose an augmentation in the latent space of the decoder to improve its robustness. Concretely, we create virtual cases by manifold softmixing two decoder hidden states of similar semantic meanings. Experimental results on Gigaword and CNN/DM datasets demonstrate that our approach achieves significant improvements over strong baselines and exhibits higher robustness on noisy, attacked, and clean datasets | [
"Chen, Xiuying",
"Long, Guodong",
"Tao, Chongyang",
"Li, Mingzhe",
"Gao, Xin",
"Zhang, Chengqi",
"Zhang, Xiangliang"
] | Improving the Robustness of Summarization Systems with Dual Augmentation | acl-long.378 | Poster | 2306.01090 | [
"https://github.com/iriscxy/robustness"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.379.bib | https://aclanthology.org/2023.acl-long.379/ | @inproceedings{zhang-etal-2023-interpretable,
title = "Interpretable Math Word Problem Solution Generation via Step-by-step Planning",
author = "Zhang, Mengxue and
Wang, Zichao and
Yang, Zhichao and
Feng, Weiqi and
Lan, 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.379",
doi = "10.18653/v1/2023.acl-long.379",
pages = "6858--6877",
abstract = "Solutions to math word problems (MWPs) with step-by-step explanations are valuable, especially in education, to help students better comprehend problem-solving strategies. Most existing approaches only focus on obtaining the final correct answer. A few recent approaches leverage intermediate solution steps to improve final answer correctness but often cannot generate coherent steps with a clear solution strategy. Contrary to existing work, we focus on improving the correctness and coherence of the intermediate solutions steps. We propose a step-by-step planning approach for intermediate solution generation, which strategically plans the generation of the next solution step based on the MWP and the previous solution steps. Our approach first plans the next step by predicting the necessary math operation needed to proceed, given history steps, then generates the next step, token-by-token, by prompting a language model with the predicted math operation. Experiments on the GSM8K dataset demonstrate that our approach improves the accuracy and interpretability of the solution on both automatic metrics and human evaluation.",
}
| Solutions to math word problems (MWPs) with step-by-step explanations are valuable, especially in education, to help students better comprehend problem-solving strategies. Most existing approaches only focus on obtaining the final correct answer. A few recent approaches leverage intermediate solution steps to improve final answer correctness but often cannot generate coherent steps with a clear solution strategy. Contrary to existing work, we focus on improving the correctness and coherence of the intermediate solutions steps. We propose a step-by-step planning approach for intermediate solution generation, which strategically plans the generation of the next solution step based on the MWP and the previous solution steps. Our approach first plans the next step by predicting the necessary math operation needed to proceed, given history steps, then generates the next step, token-by-token, by prompting a language model with the predicted math operation. Experiments on the GSM8K dataset demonstrate that our approach improves the accuracy and interpretability of the solution on both automatic metrics and human evaluation. | [
"Zhang, Mengxue",
"Wang, Zichao",
"Yang, Zhichao",
"Feng, Weiqi",
"Lan, Andrew"
] | Interpretable Math Word Problem Solution Generation via Step-by-step Planning | acl-long.379 | Poster | 2306.00784 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.380.bib | https://aclanthology.org/2023.acl-long.380/ | @inproceedings{li-etal-2023-templategec,
title = "{T}emplate{GEC}: Improving Grammatical Error Correction with Detection Template",
author = "Li, Yinghao and
Liu, Xuebo and
Wang, Shuo and
Gong, Peiyuan and
Wong, Derek F. and
Gao, Yang and
Huang, Heyan 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.380",
doi = "10.18653/v1/2023.acl-long.380",
pages = "6878--6892",
abstract = "Grammatical error correction (GEC) can be divided into sequence-to-edit (Seq2Edit) and sequence-to-sequence (Seq2Seq) frameworks, both of which have their pros and cons. To utilize the strengths and make up for the shortcomings of these frameworks, this paper proposes a novel method, TemplateGEC, which capitalizes on the capabilities of both Seq2Edit and Seq2Seq frameworks in error detection and correction respectively. TemplateGEC utilizes the detection labels from a Seq2Edit model, to construct the template as the input. A Seq2Seq model is employed to enforce consistency between the predictions of different templates by utilizing consistency learning. Experimental results on the Chinese NLPCC18, English BEA19 and CoNLL14 benchmarks show the effectiveness and robustness of TemplateGEC.Further analysis reveals the potential of our method in performing human-in-the-loop GEC. Source code and scripts are available at \url{https://github.com/li-aolong/TemplateGEC}.",
}
| Grammatical error correction (GEC) can be divided into sequence-to-edit (Seq2Edit) and sequence-to-sequence (Seq2Seq) frameworks, both of which have their pros and cons. To utilize the strengths and make up for the shortcomings of these frameworks, this paper proposes a novel method, TemplateGEC, which capitalizes on the capabilities of both Seq2Edit and Seq2Seq frameworks in error detection and correction respectively. TemplateGEC utilizes the detection labels from a Seq2Edit model, to construct the template as the input. A Seq2Seq model is employed to enforce consistency between the predictions of different templates by utilizing consistency learning. Experimental results on the Chinese NLPCC18, English BEA19 and CoNLL14 benchmarks show the effectiveness and robustness of TemplateGEC.Further analysis reveals the potential of our method in performing human-in-the-loop GEC. Source code and scripts are available at \url{https://github.com/li-aolong/TemplateGEC}. | [
"Li, Yinghao",
"Liu, Xuebo",
"Wang, Shuo",
"Gong, Peiyuan",
"Wong, Derek F.",
"Gao, Yang",
"Huang, Heyan",
"Zhang, Min"
] | TemplateGEC: Improving Grammatical Error Correction with Detection Template | acl-long.380 | Oral | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.381.bib | https://aclanthology.org/2023.acl-long.381/ | @inproceedings{park-park-2023-deep,
title = "Deep Model Compression Also Helps Models Capture Ambiguity",
author = "Park, Hancheol and
Park, Jong",
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.381",
doi = "10.18653/v1/2023.acl-long.381",
pages = "6893--6905",
abstract = "Natural language understanding (NLU) tasks face a non-trivial amount of ambiguous samples where veracity of their labels is debatable among annotators. NLU models should thus account for such ambiguity, but they approximate the human opinion distributions quite poorly and tend to produce over-confident predictions. To address this problem, we must consider how to exactly capture the degree of relationship between each sample and its candidate classes. In this work, we propose a novel method with deep model compression and show how such relationship can be accounted for. We see that more reasonably represented relationships can be discovered in the lower layers and that validation accuracies are converging at these layers, which naturally leads to layer pruning. We also see that distilling the relationship knowledge from a lower layer helps models produce better distribution. Experimental results demonstrate that our method makes substantial improvement on quantifying ambiguity without gold distribution labels. As positive side-effects, our method is found to reduce the model size significantly and improve latency, both attractive aspects of NLU products.",
}
| Natural language understanding (NLU) tasks face a non-trivial amount of ambiguous samples where veracity of their labels is debatable among annotators. NLU models should thus account for such ambiguity, but they approximate the human opinion distributions quite poorly and tend to produce over-confident predictions. To address this problem, we must consider how to exactly capture the degree of relationship between each sample and its candidate classes. In this work, we propose a novel method with deep model compression and show how such relationship can be accounted for. We see that more reasonably represented relationships can be discovered in the lower layers and that validation accuracies are converging at these layers, which naturally leads to layer pruning. We also see that distilling the relationship knowledge from a lower layer helps models produce better distribution. Experimental results demonstrate that our method makes substantial improvement on quantifying ambiguity without gold distribution labels. As positive side-effects, our method is found to reduce the model size significantly and improve latency, both attractive aspects of NLU products. | [
"Park, Hancheol",
"Park, Jong"
] | Deep Model Compression Also Helps Models Capture Ambiguity | acl-long.381 | Poster | 2306.07061 | [
""
] | https://huggingface.co/papers/2306.07061 | 1 | 0 | 0 | 2 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.382.bib | https://aclanthology.org/2023.acl-long.382/ | @inproceedings{wu-etal-2023-experts,
title = "Are Experts Needed? On Human Evaluation of Counselling Reflection Generation",
author = "Wu, Zixiu and
Balloccu, Simone and
Reiter, Ehud and
Helaoui, Rim and
Reforgiato Recupero, Diego and
Riboni, Daniele",
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.382",
doi = "10.18653/v1/2023.acl-long.382",
pages = "6906--6930",
abstract = "Reflection is a crucial counselling skill where the therapist conveys to the client their interpretation of what the client said. Language models have recently been used to generate reflections automatically, but human evaluation is challenging, particularly due to the cost of hiring experts. Laypeople-based evaluation is less expensive and easier to scale, but its quality is unknown for reflections. Therefore, we explore whether laypeople can be an alternative to experts in evaluating a fundamental quality aspect: coherence and context-consistency. We do so by asking a group of laypeople and a group of experts to annotate both synthetic reflections and human reflections from actual therapists. We find that both laypeople and experts are reliable annotators and that they have moderate-to-strong inter-group correlation, which shows that laypeople can be trusted for such evaluations. We also discover that GPT-3 mostly produces coherent and consistent reflections, and we explore changes in evaluation results when the source of synthetic reflections changes to GPT-3 from the less powerful GPT-2.",
}
| Reflection is a crucial counselling skill where the therapist conveys to the client their interpretation of what the client said. Language models have recently been used to generate reflections automatically, but human evaluation is challenging, particularly due to the cost of hiring experts. Laypeople-based evaluation is less expensive and easier to scale, but its quality is unknown for reflections. Therefore, we explore whether laypeople can be an alternative to experts in evaluating a fundamental quality aspect: coherence and context-consistency. We do so by asking a group of laypeople and a group of experts to annotate both synthetic reflections and human reflections from actual therapists. We find that both laypeople and experts are reliable annotators and that they have moderate-to-strong inter-group correlation, which shows that laypeople can be trusted for such evaluations. We also discover that GPT-3 mostly produces coherent and consistent reflections, and we explore changes in evaluation results when the source of synthetic reflections changes to GPT-3 from the less powerful GPT-2. | [
"Wu, Zixiu",
"Balloccu, Simone",
"Reiter, Ehud",
"Helaoui, Rim",
"Reforgiato Recupero, Diego",
"Riboni, Daniele"
] | Are Experts Needed? On Human Evaluation of Counselling Reflection Generation | acl-long.382 | Oral | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.383.bib | https://aclanthology.org/2023.acl-long.383/ | @inproceedings{kobayashi-etal-2023-pairspanbert,
title = "{P}air{S}pan{BERT}: An Enhanced Language Model for Bridging Resolution",
author = "Kobayashi, Hideo and
Hou, Yufang and
Ng, Vincent",
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.383",
doi = "10.18653/v1/2023.acl-long.383",
pages = "6931--6946",
abstract = "We present PairSpanBERT, a SpanBERT-based pre-trained model specialized for bridging resolution. To this end, we design a novel pre-training objective that aims to learn the contexts in which two mentions are implicitly linked to each other from a large amount of data automatically generated either heuristically or via distance supervision with a knowledge graph. Despite the noise inherent in the automatically generated data, we achieve the best results reported to date on three evaluation datasets for bridging resolution when replacing SpanBERT with PairSpanBERT in a state-of-the-art resolver that jointly performs entity coreference resolution and bridging resolution.",
}
| We present PairSpanBERT, a SpanBERT-based pre-trained model specialized for bridging resolution. To this end, we design a novel pre-training objective that aims to learn the contexts in which two mentions are implicitly linked to each other from a large amount of data automatically generated either heuristically or via distance supervision with a knowledge graph. Despite the noise inherent in the automatically generated data, we achieve the best results reported to date on three evaluation datasets for bridging resolution when replacing SpanBERT with PairSpanBERT in a state-of-the-art resolver that jointly performs entity coreference resolution and bridging resolution. | [
"Kobayashi, Hideo",
"Hou, Yufang",
"Ng, Vincent"
] | PairSpanBERT: An Enhanced Language Model for Bridging Resolution | acl-long.383 | Oral | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.384.bib | https://aclanthology.org/2023.acl-long.384/ | @inproceedings{ge-etal-2023-compounding,
title = "Compounding Geometric Operations for Knowledge Graph Completion",
author = "Ge, Xiou and
Wang, Yun Cheng and
Wang, Bin and
Kuo, C.-C. Jay",
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.384",
doi = "10.18653/v1/2023.acl-long.384",
pages = "6947--6965",
abstract = "Geometric transformations including translation, rotation, and scaling are commonly used operations in image processing. Besides, some of them are successfully used in developing effective knowledge graph embedding (KGE). Inspired by the synergy, we propose a new KGE model by leveraging all three operations in this work. Since translation, rotation, and scaling operations are cascaded to form a composite one, the new model is named CompoundE. By casting CompoundE in the framework of group theory, we show that quite a few distanced-based KGE models are special cases of CompoundE. CompoundE extends the simple distance-based scoring functions to relation-dependent compound operations on head and/or tail entities. To demonstrate the effectiveness of CompoundE, we perform three prevalent KG prediction tasks including link prediction, path query answering, and entity typing, on a range of datasets. CompoundE outperforms extant models consistently, demonstrating its effectiveness and flexibility.",
}
| Geometric transformations including translation, rotation, and scaling are commonly used operations in image processing. Besides, some of them are successfully used in developing effective knowledge graph embedding (KGE). Inspired by the synergy, we propose a new KGE model by leveraging all three operations in this work. Since translation, rotation, and scaling operations are cascaded to form a composite one, the new model is named CompoundE. By casting CompoundE in the framework of group theory, we show that quite a few distanced-based KGE models are special cases of CompoundE. CompoundE extends the simple distance-based scoring functions to relation-dependent compound operations on head and/or tail entities. To demonstrate the effectiveness of CompoundE, we perform three prevalent KG prediction tasks including link prediction, path query answering, and entity typing, on a range of datasets. CompoundE outperforms extant models consistently, demonstrating its effectiveness and flexibility. | [
"Ge, Xiou",
"Wang, Yun Cheng",
"Wang, Bin",
"Kuo, C.-C. Jay"
] | Compounding Geometric Operations for Knowledge Graph Completion | acl-long.384 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.385.bib | https://aclanthology.org/2023.acl-long.385/ | @inproceedings{li-etal-2023-shot,
title = "Few-shot In-context Learning on Knowledge Base Question Answering",
author = "Li, Tianle and
Ma, Xueguang and
Zhuang, Alex and
Gu, Yu and
Su, Yu and
Chen, Wenhu",
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.385",
doi = "10.18653/v1/2023.acl-long.385",
pages = "6966--6980",
abstract = "Question answering over knowledge bases is considered a difficult problem due to the challenge of generalizing to a wide variety of possible natural language questions. Additionally, the heterogeneity of knowledge base schema items between different knowledge bases often necessitates specialized training for different knowledge base question-answering (KBQA) datasets. To handle questions over diverse KBQA datasets with a unified training-free framework, we propose KB-BINDER, which for the first time enables few-shot in-context learning over KBQA tasks. Firstly, KB-BINDER leverages large language models like Codex to generate logical forms as the draft for a specific question by imitating a few demonstrations. Secondly, KB-BINDER grounds on the knowledge base to bind the generated draft to an executable one with BM25 score matching. The experimental results on four public heterogeneous KBQA datasets show that KB-BINDER can achieve a strong performance with only a few in-context demonstrations. Especially on GraphQA and 3-hop MetaQA, KB-BINDER can even outperform the state-of-the-art trained models. On GrailQA and WebQSP, our model is also on par with other fully-trained models. We believe KB-BINDER can serve as an important baseline for future research. We plan to release all the code and data. Our code is available at \url{https://github.com/ltl3A87/KB-BINDER}.",
}
| Question answering over knowledge bases is considered a difficult problem due to the challenge of generalizing to a wide variety of possible natural language questions. Additionally, the heterogeneity of knowledge base schema items between different knowledge bases often necessitates specialized training for different knowledge base question-answering (KBQA) datasets. To handle questions over diverse KBQA datasets with a unified training-free framework, we propose KB-BINDER, which for the first time enables few-shot in-context learning over KBQA tasks. Firstly, KB-BINDER leverages large language models like Codex to generate logical forms as the draft for a specific question by imitating a few demonstrations. Secondly, KB-BINDER grounds on the knowledge base to bind the generated draft to an executable one with BM25 score matching. The experimental results on four public heterogeneous KBQA datasets show that KB-BINDER can achieve a strong performance with only a few in-context demonstrations. Especially on GraphQA and 3-hop MetaQA, KB-BINDER can even outperform the state-of-the-art trained models. On GrailQA and WebQSP, our model is also on par with other fully-trained models. We believe KB-BINDER can serve as an important baseline for future research. We plan to release all the code and data. Our code is available at \url{https://github.com/ltl3A87/KB-BINDER}. | [
"Li, Tianle",
"Ma, Xueguang",
"Zhuang, Alex",
"Gu, Yu",
"Su, Yu",
"Chen, Wenhu"
] | Few-shot In-context Learning on Knowledge Base Question Answering | acl-long.385 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.386.bib | https://aclanthology.org/2023.acl-long.386/ | @inproceedings{pan-etal-2023-fact,
title = "Fact-Checking Complex Claims with Program-Guided Reasoning",
author = "Pan, Liangming and
Wu, Xiaobao and
Lu, Xinyuan and
Luu, Anh Tuan and
Wang, William Yang and
Kan, Min-Yen and
Nakov, Preslav",
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.386",
doi = "10.18653/v1/2023.acl-long.386",
pages = "6981--7004",
abstract = "Fact-checking real-world claims often requires collecting multiple pieces of evidence and applying complex multi-step reasoning. In this paper, we present Program-Guided Fact-Checking (ProgramFC), a novel fact-checking model that decomposes complex claims into simpler sub-tasks that can be solved using a shared library of specialized functions. We first leverage the in-context learning ability of large language models to generate reasoning programs to guide the verification process. Afterward, we execute the program by delegating each sub-task to the corresponding sub-task handler. This process makes our model both explanatory and data-efficient, providing clear explanations of its reasoning process and requiring minimal training data. We evaluate ProgramFC on two challenging fact-checking datasets and show that it outperforms seven fact-checking baselines across different settings of evidence availability, with explicit output programs that benefit human debugging. Our codes and data are publicly available at \url{https://github.com/mbzuai-nlp/ProgramFC}.",
}
| Fact-checking real-world claims often requires collecting multiple pieces of evidence and applying complex multi-step reasoning. In this paper, we present Program-Guided Fact-Checking (ProgramFC), a novel fact-checking model that decomposes complex claims into simpler sub-tasks that can be solved using a shared library of specialized functions. We first leverage the in-context learning ability of large language models to generate reasoning programs to guide the verification process. Afterward, we execute the program by delegating each sub-task to the corresponding sub-task handler. This process makes our model both explanatory and data-efficient, providing clear explanations of its reasoning process and requiring minimal training data. We evaluate ProgramFC on two challenging fact-checking datasets and show that it outperforms seven fact-checking baselines across different settings of evidence availability, with explicit output programs that benefit human debugging. Our codes and data are publicly available at \url{https://github.com/mbzuai-nlp/ProgramFC}. | [
"Pan, Liangming",
"Wu, Xiaobao",
"Lu, Xinyuan",
"Luu, Anh Tuan",
"Wang, William Yang",
"Kan, Min-Yen",
"Nakov, Preslav"
] | Fact-Checking Complex Claims with Program-Guided Reasoning | acl-long.386 | Poster | 2305.12744 | [
"https://github.com/mbzuai-nlp/programfc"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.387.bib | https://aclanthology.org/2023.acl-long.387/ | @inproceedings{jin-etal-2023-patton,
title = "Patton: Language Model Pretraining on Text-Rich Networks",
author = "Jin, Bowen and
Zhang, Wentao and
Zhang, Yu and
Meng, Yu and
Zhang, Xinyang and
Zhu, Qi and
Han, Jiawei",
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.387",
doi = "10.18653/v1/2023.acl-long.387",
pages = "7005--7020",
abstract = "A real-world text corpus sometimes comprises not only text documents, but also semantic links between them (e.g., academic papers in a bibliographic network are linked by citations and co-authorships).Text documents and semantic connections form a text-rich network, which empowers a wide range of downstream tasks such as classification and retrieval. However, pretraining methods for such structures are still lacking, making it difficult to build one generic model that can be adapted to various tasks on text-rich networks. Current pretraining objectives, such as masked language modeling, purely model texts and do not take inter-document structure information into consideration. To this end, we propose our PretrAining on TexT-Rich NetwOrk framework Patton.Patton includes two pretraining strategies: network-contextualized masked language modeling and masked node prediction, to capture the inherent dependency between textual attributes and network structure. We conduct experiments on four downstream tasks in five datasets from both academic and e-commerce domains, where Patton outperforms baselines significantly and consistently.",
}
| A real-world text corpus sometimes comprises not only text documents, but also semantic links between them (e.g., academic papers in a bibliographic network are linked by citations and co-authorships).Text documents and semantic connections form a text-rich network, which empowers a wide range of downstream tasks such as classification and retrieval. However, pretraining methods for such structures are still lacking, making it difficult to build one generic model that can be adapted to various tasks on text-rich networks. Current pretraining objectives, such as masked language modeling, purely model texts and do not take inter-document structure information into consideration. To this end, we propose our PretrAining on TexT-Rich NetwOrk framework Patton.Patton includes two pretraining strategies: network-contextualized masked language modeling and masked node prediction, to capture the inherent dependency between textual attributes and network structure. We conduct experiments on four downstream tasks in five datasets from both academic and e-commerce domains, where Patton outperforms baselines significantly and consistently. | [
"Jin, Bowen",
"Zhang, Wentao",
"Zhang, Yu",
"Meng, Yu",
"Zhang, Xinyang",
"Zhu, Qi",
"Han, Jiawei"
] | Patton: Language Model Pretraining on Text-Rich Networks | acl-long.387 | Oral | 2305.12268 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.388.bib | https://aclanthology.org/2023.acl-long.388/ | @inproceedings{zeng-etal-2023-soft,
title = "Soft Language Clustering for Multilingual Model Pre-training",
author = "Zeng, Jiali and
Jiang, Yufan and
Yin, Yongjing and
Jing, Yi and
Meng, Fandong and
Lin, Binghuai and
Cao, Yunbo and
Zhou, Jie",
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.388",
doi = "10.18653/v1/2023.acl-long.388",
pages = "7021--7035",
abstract = "Multilingual pre-trained language models have demonstrated impressive (zero-shot) cross-lingual transfer abilities, however, their performance is hindered when the target language has distant typologyfrom the source language or when pre-training data is limited in size. In this paper, we propose XLM-P, a method that contextually retrieves prompts as flexible guidance for encoding instances conditionally. Our space-efficient and model-agnostic XLM-P approach enables (1) lightweight modeling of language-invariant and language-specific knowledge across languages, and (2) easy integration with other multilingual pre-training methods. On the tasks of XTREME, which include text classification, sequence labeling, question answering, and sentence retrieval, both base- and large-size language models pre-trained with our proposed method exhibit consistent performance improvement. Furthermore, it provides substantial advantages for low-resource languages in unsupervised sentence retrieval and for target languages that differ greatly from the source language in cross-lingual transfer.",
}
| Multilingual pre-trained language models have demonstrated impressive (zero-shot) cross-lingual transfer abilities, however, their performance is hindered when the target language has distant typologyfrom the source language or when pre-training data is limited in size. In this paper, we propose XLM-P, a method that contextually retrieves prompts as flexible guidance for encoding instances conditionally. Our space-efficient and model-agnostic XLM-P approach enables (1) lightweight modeling of language-invariant and language-specific knowledge across languages, and (2) easy integration with other multilingual pre-training methods. On the tasks of XTREME, which include text classification, sequence labeling, question answering, and sentence retrieval, both base- and large-size language models pre-trained with our proposed method exhibit consistent performance improvement. Furthermore, it provides substantial advantages for low-resource languages in unsupervised sentence retrieval and for target languages that differ greatly from the source language in cross-lingual transfer. | [
"Zeng, Jiali",
"Jiang, Yufan",
"Yin, Yongjing",
"Jing, Yi",
"Meng, F",
"ong",
"Lin, Binghuai",
"Cao, Yunbo",
"Zhou, Jie"
] | Soft Language Clustering for Multilingual Model Pre-training | acl-long.388 | Poster | 2306.07610 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.389.bib | https://aclanthology.org/2023.acl-long.389/ | @inproceedings{vakil-amiri-2023-curriculum,
title = "Curriculum Learning for Graph Neural Networks: A Multiview Competence-based Approach",
author = "Vakil, Nidhi and
Amiri, Hadi",
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.389",
doi = "10.18653/v1/2023.acl-long.389",
pages = "7036--7051",
abstract = "A curriculum is a planned sequence of learning materials and an effective one can make learning efficient and effective for both humans and machines. Recent studies developed effective data-driven curriculum learning approaches for training graph neural networks in language applications. However, existing curriculum learning approaches often employ a single criterion of difficulty in their training paradigms. In this paper, we propose a new perspective on curriculum learning by introducing a novel approach that builds on graph complexity formalisms (as difficulty criteria) and model competence during training. The model consists of a scheduling scheme which derives effective curricula by accounting for different views of sample difficulty and model competence during training. The proposed solution advances existing research in curriculum learning for graph neural networks with the ability to incorporate a fine-grained spectrum of graph difficulty criteria in their training paradigms. Experimental results on real-world link prediction and node classification tasks illustrate the effectiveness of the proposed approach.",
}
| A curriculum is a planned sequence of learning materials and an effective one can make learning efficient and effective for both humans and machines. Recent studies developed effective data-driven curriculum learning approaches for training graph neural networks in language applications. However, existing curriculum learning approaches often employ a single criterion of difficulty in their training paradigms. In this paper, we propose a new perspective on curriculum learning by introducing a novel approach that builds on graph complexity formalisms (as difficulty criteria) and model competence during training. The model consists of a scheduling scheme which derives effective curricula by accounting for different views of sample difficulty and model competence during training. The proposed solution advances existing research in curriculum learning for graph neural networks with the ability to incorporate a fine-grained spectrum of graph difficulty criteria in their training paradigms. Experimental results on real-world link prediction and node classification tasks illustrate the effectiveness of the proposed approach. | [
"Vakil, Nidhi",
"Amiri, Hadi"
] | Curriculum Learning for Graph Neural Networks: A Multiview Competence-based Approach | acl-long.389 | Poster | 2307.08859 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.390.bib | https://aclanthology.org/2023.acl-long.390/ | @inproceedings{sharma-etal-2023-paraphrase,
title = "When and how to paraphrase for named entity recognition?",
author = "Sharma, Saket and
Joshi, Aviral and
Zhao, Yiyun and
Mukhija, Namrata and
Bhathena, Hanoz and
Singh, Prateek and
Santhanam, Sashank",
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.390",
doi = "10.18653/v1/2023.acl-long.390",
pages = "7052--7087",
abstract = "While paraphrasing is a promising approach for data augmentation in classification tasks, its effect on named entity recognition (NER) is not investigated systematically due to the difficulty of span-level label preservation. In this paper, we utilize simple strategies to annotate entity spans in generations and compare established and novel methods of paraphrasing in NLP such as back translation, specialized encoder-decoder models such as Pegasus, and GPT-3 variants for their effectiveness in improving downstream performance for NER across different levels of gold annotations and paraphrasing strength on 5 datasets. We thoroughly explore the influence of paraphrasers, and dynamics between paraphrasing strength and gold dataset size on the NER performance with visualizations and statistical testing. We find that the choice of the paraphraser greatly impacts NER performance, with one of the larger GPT-3 variants exceedingly capable of generating high quality paraphrases, yielding statistically significant improvements in NER performance with increasing paraphrasing strength, while other paraphrasers show more mixed results. Additionally, inline auto annotations generated by larger GPT-3 are strictly better than heuristic based annotations. We also find diminishing benefits of paraphrasing as gold annotations increase for most datasets. Furthermore, while most paraphrasers promote entity memorization in NER, the proposed GPT-3 configuration performs most favorably among the compared paraphrasers when tested on unseen entities, with memorization reducing further with paraphrasing strength. Finally, we explore mention replacement using GPT-3, which provides additional benefits over base paraphrasing for specific datasets.",
}
| While paraphrasing is a promising approach for data augmentation in classification tasks, its effect on named entity recognition (NER) is not investigated systematically due to the difficulty of span-level label preservation. In this paper, we utilize simple strategies to annotate entity spans in generations and compare established and novel methods of paraphrasing in NLP such as back translation, specialized encoder-decoder models such as Pegasus, and GPT-3 variants for their effectiveness in improving downstream performance for NER across different levels of gold annotations and paraphrasing strength on 5 datasets. We thoroughly explore the influence of paraphrasers, and dynamics between paraphrasing strength and gold dataset size on the NER performance with visualizations and statistical testing. We find that the choice of the paraphraser greatly impacts NER performance, with one of the larger GPT-3 variants exceedingly capable of generating high quality paraphrases, yielding statistically significant improvements in NER performance with increasing paraphrasing strength, while other paraphrasers show more mixed results. Additionally, inline auto annotations generated by larger GPT-3 are strictly better than heuristic based annotations. We also find diminishing benefits of paraphrasing as gold annotations increase for most datasets. Furthermore, while most paraphrasers promote entity memorization in NER, the proposed GPT-3 configuration performs most favorably among the compared paraphrasers when tested on unseen entities, with memorization reducing further with paraphrasing strength. Finally, we explore mention replacement using GPT-3, which provides additional benefits over base paraphrasing for specific datasets. | [
"Sharma, Saket",
"Joshi, Aviral",
"Zhao, Yiyun",
"Mukhija, Namrata",
"Bhathena, Hanoz",
"Singh, Prateek",
"Santhanam, Sashank"
] | When and how to paraphrase for named entity recognition? | acl-long.390 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.391.bib | https://aclanthology.org/2023.acl-long.391/ | @inproceedings{tao-etal-2023-unievent,
title = "{U}ni{E}vent: Unified Generative Model with Multi-Dimensional Prefix for Zero-Shot Event-Relational Reasoning",
author = "Tao, Zhengwei and
Jin, Zhi and
Zhao, Haiyan and
Dou, Chengfeng and
Zhao, Yongqiang and
Shen, Tao and
Tao, Chongyang",
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.391",
doi = "10.18653/v1/2023.acl-long.391",
pages = "7088--7102",
abstract = "Reasoning about events and their relations attracts surging research efforts since it is regarded as an indispensable ability to fulfill various event-centric or common-sense reasoning tasks. However, these tasks often suffer from limited data availability due to the labor-intensive nature of their annotations. Consequently, recent studies have explored knowledge transfer approaches within a multi-task learning framework to address this challenge. Although such methods have achieved acceptable results, such brute-force solutions struggle to effectively transfer event-relational knowledge due to the vast array of inter-event relations (e.g. temporal, causal, conditional) and reasoning formulations (e.g. discriminative, abductive, ending prediction). To enhance knowledge transfer and enable zero-shot generalization among various combinations, in this work we propose a novel unified framework, called UNIEVENT. Inspired by prefix-based multitask learning, our approach organizes event relational reasoning tasks into a coordinate system with multiple axes, representing inter-event relations and reasoning formulations. We then train a unified text-to-text generative model that utilizes coordinate-assigning prefixes for each task. By leveraging our adapted prefixes, our unified model achieves state-of-the-art or competitive performance on both zero-shot and supervised reasoning tasks, as demonstrated in extensive experiments",
}
| Reasoning about events and their relations attracts surging research efforts since it is regarded as an indispensable ability to fulfill various event-centric or common-sense reasoning tasks. However, these tasks often suffer from limited data availability due to the labor-intensive nature of their annotations. Consequently, recent studies have explored knowledge transfer approaches within a multi-task learning framework to address this challenge. Although such methods have achieved acceptable results, such brute-force solutions struggle to effectively transfer event-relational knowledge due to the vast array of inter-event relations (e.g. temporal, causal, conditional) and reasoning formulations (e.g. discriminative, abductive, ending prediction). To enhance knowledge transfer and enable zero-shot generalization among various combinations, in this work we propose a novel unified framework, called UNIEVENT. Inspired by prefix-based multitask learning, our approach organizes event relational reasoning tasks into a coordinate system with multiple axes, representing inter-event relations and reasoning formulations. We then train a unified text-to-text generative model that utilizes coordinate-assigning prefixes for each task. By leveraging our adapted prefixes, our unified model achieves state-of-the-art or competitive performance on both zero-shot and supervised reasoning tasks, as demonstrated in extensive experiments | [
"Tao, Zhengwei",
"Jin, Zhi",
"Zhao, Haiyan",
"Dou, Chengfeng",
"Zhao, Yongqiang",
"Shen, Tao",
"Tao, Chongyang"
] | UniEvent: Unified Generative Model with Multi-Dimensional Prefix for Zero-Shot Event-Relational Reasoning | acl-long.391 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2023.acl-long.392.bib | https://aclanthology.org/2023.acl-long.392/ | @inproceedings{joshi-etal-2023-machine,
title = "Are Machine Rationales (Not) Useful to Humans? Measuring and Improving Human Utility of Free-text Rationales",
author = "Joshi, Brihi and
Liu, Ziyi and
Ramnath, Sahana and
Chan, Aaron and
Tong, Zhewei and
Nie, Shaoliang and
Wang, Qifan and
Choi, Yejin and
Ren, Xiang",
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.392",
doi = "10.18653/v1/2023.acl-long.392",
pages = "7103--7128",
abstract = "Among the remarkable emergent capabilities of large language models (LMs) is free-text rationalization; beyond certain scale, large LMs are capable of generating seemingly useful rationalizations, which in turn, can dramatically enhance their performances on leaderboards. This phenomenon raises a question: can machine generated rationales also be useful for humans, especially when lay humans try to answer questions based on those machine rationales? We observe that human utility of existing rationales is far from satisfactory and expensive to estimate with human studies. Existing metrics like task performance of the LM generating the rationales or similarity between generated and gold rationales are not good indicators of their human utility. While we observe that certain properties of rationales like conciseness and novelty are correlated with their human utility, estimating them without human involvement is challenging. We show that, by estimating a rationale{'}s helpfulness in answering similar unseen instances, we can measure its human utility to a better extent. We also translate this finding into an automated score, Gen-U, that we propose, which can help improve LMs{'} ability to generate rationales with better human utility, while maintaining most of its task performance. Lastly, we release all code and collected data with this project.",
}
| Among the remarkable emergent capabilities of large language models (LMs) is free-text rationalization; beyond certain scale, large LMs are capable of generating seemingly useful rationalizations, which in turn, can dramatically enhance their performances on leaderboards. This phenomenon raises a question: can machine generated rationales also be useful for humans, especially when lay humans try to answer questions based on those machine rationales? We observe that human utility of existing rationales is far from satisfactory and expensive to estimate with human studies. Existing metrics like task performance of the LM generating the rationales or similarity between generated and gold rationales are not good indicators of their human utility. While we observe that certain properties of rationales like conciseness and novelty are correlated with their human utility, estimating them without human involvement is challenging. We show that, by estimating a rationale{'}s helpfulness in answering similar unseen instances, we can measure its human utility to a better extent. We also translate this finding into an automated score, Gen-U, that we propose, which can help improve LMs{'} ability to generate rationales with better human utility, while maintaining most of its task performance. Lastly, we release all code and collected data with this project. | [
"Joshi, Brihi",
"Liu, Ziyi",
"Ramnath, Sahana",
"Chan, Aaron",
"Tong, Zhewei",
"Nie, Shaoliang",
"Wang, Qifan",
"Choi, Yejin",
"Ren, Xiang"
] | Are Machine Rationales (Not) Useful to Humans? Measuring and Improving Human Utility of Free-text Rationales | acl-long.392 | Oral | 2305.07095 | [
"https://github.com/ink-usc/rationalehumanutility"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.393.bib | https://aclanthology.org/2023.acl-long.393/ | @inproceedings{durandard-etal-2023-automatic,
title = "Automatic Annotation of Direct Speech in Written {F}rench Narratives",
author = "Durandard, No{\'e} and
Tran, Viet Anh and
Michel, Gaspard and
Epure, Elena",
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.393",
doi = "10.18653/v1/2023.acl-long.393",
pages = "7129--7147",
abstract = "The automatic annotation of direct speech (AADS) in written text has been often used in computational narrative understanding. Methods based on either rules or deep neural networks have been explored, in particular for English or German languages. Yet, for French, our target language, not many works exist. Our goal is to create a unified framework to design and evaluate AADS models in French. For this, we consolidated the largest-to-date French narrative dataset annotated with DS per word; we adapted various baselines for sequence labelling or from AADS in other languages; and we designed and conducted an extensive evaluation focused on generalisation. Results show that the task still requires substantial efforts and emphasise characteristics of each baseline. Although this framework could be improved, it is a step further to encourage more research on the topic.",
}
| The automatic annotation of direct speech (AADS) in written text has been often used in computational narrative understanding. Methods based on either rules or deep neural networks have been explored, in particular for English or German languages. Yet, for French, our target language, not many works exist. Our goal is to create a unified framework to design and evaluate AADS models in French. For this, we consolidated the largest-to-date French narrative dataset annotated with DS per word; we adapted various baselines for sequence labelling or from AADS in other languages; and we designed and conducted an extensive evaluation focused on generalisation. Results show that the task still requires substantial efforts and emphasise characteristics of each baseline. Although this framework could be improved, it is a step further to encourage more research on the topic. | [
"Dur",
"ard, No{\\'e}",
"Tran, Viet Anh",
"Michel, Gaspard",
"Epure, Elena"
] | Automatic Annotation of Direct Speech in Written French Narratives | acl-long.393 | Poster | 2306.15634 | [
"https://github.com/deezer/aads_french"
] | https://huggingface.co/papers/2306.15634 | 2 | 0 | 0 | 4 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.394.bib | https://aclanthology.org/2023.acl-long.394/ | @inproceedings{kim-etal-2023-automatic,
title = "Automatic Creation of Named Entity Recognition Datasets by Querying Phrase Representations",
author = "Kim, Hyunjae and
Yoo, Jaehyo and
Yoon, Seunghyun and
Kang, Jaewoo",
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.394",
doi = "10.18653/v1/2023.acl-long.394",
pages = "7148--7163",
abstract = "Most weakly supervised named entity recognition (NER) models rely on domain-specific dictionaries provided by experts. This approach is infeasible in many domains where dictionaries do not exist. While a phrase retrieval model was used to construct pseudo-dictionaries with entities retrieved from Wikipedia automatically in a recent study, these dictionaries often have limited coverage because the retriever is likely to retrieve popular entities rather than rare ones. In this study, we present a novel framework, HighGEN, that generates NER datasets with high-coverage pseudo-dictionaries. Specifically, we create entity-rich dictionaries with a novel search method, called phrase embedding search, which encourages the retriever to search a space densely populated with various entities. In addition, we use a new verification process based on the embedding distance between candidate entity mentions and entity types to reduce the false-positive noise in weak labels generated by high-coverage dictionaries. We demonstrate that HighGEN outperforms the previous best model by an average F1 score of 4.7 across five NER benchmark datasets.",
}
| Most weakly supervised named entity recognition (NER) models rely on domain-specific dictionaries provided by experts. This approach is infeasible in many domains where dictionaries do not exist. While a phrase retrieval model was used to construct pseudo-dictionaries with entities retrieved from Wikipedia automatically in a recent study, these dictionaries often have limited coverage because the retriever is likely to retrieve popular entities rather than rare ones. In this study, we present a novel framework, HighGEN, that generates NER datasets with high-coverage pseudo-dictionaries. Specifically, we create entity-rich dictionaries with a novel search method, called phrase embedding search, which encourages the retriever to search a space densely populated with various entities. In addition, we use a new verification process based on the embedding distance between candidate entity mentions and entity types to reduce the false-positive noise in weak labels generated by high-coverage dictionaries. We demonstrate that HighGEN outperforms the previous best model by an average F1 score of 4.7 across five NER benchmark datasets. | [
"Kim, Hyunjae",
"Yoo, Jaehyo",
"Yoon, Seunghyun",
"Kang, Jaewoo"
] | Automatic Creation of Named Entity Recognition Datasets by Querying Phrase Representations | acl-long.394 | Poster | 2210.07586 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.395.bib | https://aclanthology.org/2023.acl-long.395/ | @inproceedings{chen-etal-2023-dynamic,
title = "Dynamic Transformers Provide a False Sense of Efficiency",
author = "Chen, Yiming and
Chen, Simin and
Li, Zexin and
Yang, Wei and
Liu, Cong and
Tan, Robby and
Li, Haizhou",
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.395",
doi = "10.18653/v1/2023.acl-long.395",
pages = "7164--7180",
abstract = "Despite much success in natural language processing (NLP), pre-trained language models typically lead to a high computational cost during inference. Multi-exit is a mainstream approach to address this issue by making a trade-off between efficiency and accuracy, where the saving of computation comes from an early exit. However, whether such saving from early-exiting is robust remains unknown. Motivated by this, we first show that directly adapting existing adversarial attack approaches targeting model accuracy cannot significantly reduce inference efficiency. To this end, we propose a simple yet effective attacking framework, SAME, a novel slowdown attack framework on multi-exit models, which is specially tailored to reduce the efficiency of the multi-exit models. By leveraging the multi-exit models{'} design characteristics, we utilize all internal predictions to guide the adversarial sample generation instead of merely considering the final prediction. Experiments on the GLUE benchmark show that SAME can effectively diminish the efficiency gain of various multi-exit models by 80{\%} on average, convincingly validating its effectiveness and generalization ability.",
}
| Despite much success in natural language processing (NLP), pre-trained language models typically lead to a high computational cost during inference. Multi-exit is a mainstream approach to address this issue by making a trade-off between efficiency and accuracy, where the saving of computation comes from an early exit. However, whether such saving from early-exiting is robust remains unknown. Motivated by this, we first show that directly adapting existing adversarial attack approaches targeting model accuracy cannot significantly reduce inference efficiency. To this end, we propose a simple yet effective attacking framework, SAME, a novel slowdown attack framework on multi-exit models, which is specially tailored to reduce the efficiency of the multi-exit models. By leveraging the multi-exit models{'} design characteristics, we utilize all internal predictions to guide the adversarial sample generation instead of merely considering the final prediction. Experiments on the GLUE benchmark show that SAME can effectively diminish the efficiency gain of various multi-exit models by 80{\%} on average, convincingly validating its effectiveness and generalization ability. | [
"Chen, Yiming",
"Chen, Simin",
"Li, Zexin",
"Yang, Wei",
"Liu, Cong",
"Tan, Robby",
"Li, Haizhou"
] | Dynamic Transformers Provide a False Sense of Efficiency | acl-long.395 | Poster | 2305.12228 | [
"https://github.com/matthewcym/same"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2023.acl-long.396.bib | https://aclanthology.org/2023.acl-long.396/ | @inproceedings{hlavnova-ruder-2023-empowering,
title = "Empowering Cross-lingual Behavioral Testing of {NLP} Models with Typological Features",
author = "Hlavnova, Ester and
Ruder, Sebastian",
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.396",
doi = "10.18653/v1/2023.acl-long.396",
pages = "7181--7198",
abstract = "A challenge towards developing NLP systems for the world{'}s languages is understanding how they generalize to typological differences relevant for real-world applications. To this end, we propose M2C, a morphologically-aware framework for behavioral testing of NLP models. We use M2C to generate tests that probe models{'} behavior in light of specific linguistic features in 12 typologically diverse languages. We evaluate state-of-the-art language models on the generated tests. While models excel at most tests in English, we highlight generalization failures to specific typological characteristics such as temporal expressions in Swahili and compounding possessives in Finish. Our findings motivate the development of models that address these blind spots.",
}
| A challenge towards developing NLP systems for the world{'}s languages is understanding how they generalize to typological differences relevant for real-world applications. To this end, we propose M2C, a morphologically-aware framework for behavioral testing of NLP models. We use M2C to generate tests that probe models{'} behavior in light of specific linguistic features in 12 typologically diverse languages. We evaluate state-of-the-art language models on the generated tests. While models excel at most tests in English, we highlight generalization failures to specific typological characteristics such as temporal expressions in Swahili and compounding possessives in Finish. Our findings motivate the development of models that address these blind spots. | [
"Hlavnova, Ester",
"Ruder, Sebastian"
] | Empowering Cross-lingual Behavioral Testing of NLP Models with Typological Features | acl-long.396 | Oral | 2307.05454 | [
"https://github.com/google-research/multi-morph-checklist"
] | https://huggingface.co/papers/2307.05454 | 0 | 6 | 0 | 2 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.397.bib | https://aclanthology.org/2023.acl-long.397/ | @inproceedings{sreedhar-etal-2023-local,
title = "Local Byte Fusion for Neural Machine Translation",
author = "Sreedhar, Makesh Narsimhan and
Wan, Xiangpeng and
Cheng, Yu and
Hu, Junjie",
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.397",
doi = "10.18653/v1/2023.acl-long.397",
pages = "7199--7214",
abstract = "Subword tokenization schemes are the dominant technique used in current NLP models. However, such schemes can be rigid and tokenizers built on one corpus may not adapt well to other parallel corpora. It has also been observed that in multilingual corpora, subword tokenization schemes oversegment low-resource languages, leading to a drop in translation performance. An alternative to subword tokenizers is byte-based tokenization, i.e., tokenization into byte sequences using the UTF-8 encoding scheme. Byte tokens often represent inputs at a sub-character granularity, i.e., one character can be represented by a span of byte tokens. This results in much longer byte sequences that are hard to interpret without aggregating local information from multiple byte tokens. In this paper, we propose a Local Byte Fusion (LOBEF) method for byte-based machine translation{---}utilizing byte n-gram and word boundaries{---}to aggregate local semantic information. Extensive experiments on multilingual translation, zero-shot cross-lingual transfer, and domain adaptation reveal a consistent improvement over vanilla byte-based models. Further analysis also indicates that our byte-based models are parameter-efficient and perform competitive to subword models.",
}
| Subword tokenization schemes are the dominant technique used in current NLP models. However, such schemes can be rigid and tokenizers built on one corpus may not adapt well to other parallel corpora. It has also been observed that in multilingual corpora, subword tokenization schemes oversegment low-resource languages, leading to a drop in translation performance. An alternative to subword tokenizers is byte-based tokenization, i.e., tokenization into byte sequences using the UTF-8 encoding scheme. Byte tokens often represent inputs at a sub-character granularity, i.e., one character can be represented by a span of byte tokens. This results in much longer byte sequences that are hard to interpret without aggregating local information from multiple byte tokens. In this paper, we propose a Local Byte Fusion (LOBEF) method for byte-based machine translation{---}utilizing byte n-gram and word boundaries{---}to aggregate local semantic information. Extensive experiments on multilingual translation, zero-shot cross-lingual transfer, and domain adaptation reveal a consistent improvement over vanilla byte-based models. Further analysis also indicates that our byte-based models are parameter-efficient and perform competitive to subword models. | [
"Sreedhar, Makesh Narsimhan",
"Wan, Xiangpeng",
"Cheng, Yu",
"Hu, Junjie"
] | Local Byte Fusion for Neural Machine Translation | acl-long.397 | Poster | 2205.11490 | [
"https://github.com/makeshn/lobef_byte_nmt"
] | https://huggingface.co/papers/2205.11490 | 0 | 0 | 0 | 4 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.398.bib | https://aclanthology.org/2023.acl-long.398/ | @inproceedings{minixhofer-etal-2023-wheres,
title = "Where{'}s the Point? Self-Supervised Multilingual Punctuation-Agnostic Sentence Segmentation",
author = "Minixhofer, Benjamin and
Pfeiffer, Jonas and
Vuli{\'c}, Ivan",
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.398",
doi = "10.18653/v1/2023.acl-long.398",
pages = "7215--7235",
abstract = "Many NLP pipelines split text into sentences as one of the crucial preprocessing steps. Prior sentence segmentation tools either rely on punctuation or require a considerable amount of sentence-segmented training data: both central assumptions might fail when porting sentence segmenters to diverse languages on a massive scale. In this work, we thus introduce a multilingual punctuation-agnostic sentence segmentation method, currently covering 85 languages, trained in a self-supervised fashion on unsegmented text, by making use of newline characters which implicitly perform segmentation into paragraphs. We further propose an approach that adapts our method to the segmentation in a given corpus by using only a small number (64-256) of sentence-segmented examples. The main results indicate that our method outperforms all the prior best sentence-segmentation tools by an average of 6.1{\%} F1 points. Furthermore, we demonstrate that proper sentence segmentation has a point: the use of a (powerful) sentence segmenter makes a considerable difference for a downstream application such as machine translation (MT). By using our method to match sentence segmentation to the segmentation used during training of MT models, we achieve an average improvement of 2.3 BLEU points over the best prior segmentation tool, as well as massive gains over a trivial segmenter that splits text into equally-sized blocks.",
}
| Many NLP pipelines split text into sentences as one of the crucial preprocessing steps. Prior sentence segmentation tools either rely on punctuation or require a considerable amount of sentence-segmented training data: both central assumptions might fail when porting sentence segmenters to diverse languages on a massive scale. In this work, we thus introduce a multilingual punctuation-agnostic sentence segmentation method, currently covering 85 languages, trained in a self-supervised fashion on unsegmented text, by making use of newline characters which implicitly perform segmentation into paragraphs. We further propose an approach that adapts our method to the segmentation in a given corpus by using only a small number (64-256) of sentence-segmented examples. The main results indicate that our method outperforms all the prior best sentence-segmentation tools by an average of 6.1{\%} F1 points. Furthermore, we demonstrate that proper sentence segmentation has a point: the use of a (powerful) sentence segmenter makes a considerable difference for a downstream application such as machine translation (MT). By using our method to match sentence segmentation to the segmentation used during training of MT models, we achieve an average improvement of 2.3 BLEU points over the best prior segmentation tool, as well as massive gains over a trivial segmenter that splits text into equally-sized blocks. | [
"Minixhofer, Benjamin",
"Pfeiffer, Jonas",
"Vuli{\\'c}, Ivan"
] | Where's the Point? Self-Supervised Multilingual Punctuation-Agnostic Sentence Segmentation | acl-long.398 | Poster | 2305.18893 | [
""
] | https://huggingface.co/papers/2305.18893 | 1 | 2 | 0 | 3 | 1 | [] | [] | [] |
https://aclanthology.org/2023.acl-long.399.bib | https://aclanthology.org/2023.acl-long.399/ | @inproceedings{li-etal-2023-multi-target,
title = "Multi-target Backdoor Attacks for Code Pre-trained Models",
author = "Li, Yanzhou and
Liu, Shangqing and
Chen, Kangjie and
Xie, Xiaofei and
Zhang, Tianwei and
Liu, 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.399",
doi = "10.18653/v1/2023.acl-long.399",
pages = "7236--7254",
abstract = "Backdoor attacks for neural code models have gained considerable attention due to the advancement of code intelligence. However, most existing works insert triggers into task-specific data for code-related downstream tasks, thereby limiting the scope of attacks. Moreover, the majority of attacks for pre-trained models are designed for understanding tasks. In this paper, we propose task-agnostic backdoor attacks for code pre-trained models. Our backdoored model is pre-trained with two learning strategies (i.e., Poisoned Seq2Seq learning and token representation learning) to support the multi-target attack of downstream code understanding and generation tasks. During the deployment phase, the implanted backdoors in the victim models can be activated by the designed triggers to achieve the targeted attack. We evaluate our approach on two code understanding tasks and three code generation tasks over seven datasets. Extensive experimental results demonstrate that our approach effectively and stealthily attacks code-related downstream tasks.",
}
| Backdoor attacks for neural code models have gained considerable attention due to the advancement of code intelligence. However, most existing works insert triggers into task-specific data for code-related downstream tasks, thereby limiting the scope of attacks. Moreover, the majority of attacks for pre-trained models are designed for understanding tasks. In this paper, we propose task-agnostic backdoor attacks for code pre-trained models. Our backdoored model is pre-trained with two learning strategies (i.e., Poisoned Seq2Seq learning and token representation learning) to support the multi-target attack of downstream code understanding and generation tasks. During the deployment phase, the implanted backdoors in the victim models can be activated by the designed triggers to achieve the targeted attack. We evaluate our approach on two code understanding tasks and three code generation tasks over seven datasets. Extensive experimental results demonstrate that our approach effectively and stealthily attacks code-related downstream tasks. | [
"Li, Yanzhou",
"Liu, Shangqing",
"Chen, Kangjie",
"Xie, Xiaofei",
"Zhang, Tianwei",
"Liu, Yang"
] | Multi-target Backdoor Attacks for Code Pre-trained Models | acl-long.399 | Poster | 2306.08350 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
Subsets and Splits