@inproceedings{yasunaga-etal-2021-lm,
title = "{LM}-Critic: Language Models for Unsupervised Grammatical Error Correction",
author = "Yasunaga, Michihiro and
Leskovec, Jure and
Liang, Percy",
booktitle = "Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2021",
address = "Online and Punta Cana, Dominican Republic",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.emnlp-main.611",
doi = "10.18653/v1/2021.emnlp-main.611",
pages = "7752--7763",
abstract = "Grammatical error correction (GEC) requires a set of labeled ungrammatical / grammatical sentence pairs for training, but obtaining such annotation can be prohibitively expensive. Recently, the Break-It-Fix-It (BIFI) framework has demonstrated strong results on learning to repair a broken program without any labeled examples, but this relies on a perfect critic (e.g., a compiler) that returns whether an example is valid or not, which does not exist for the GEC task. In this work, we show how to leverage a pretrained language model (LM) in defining an LM-Critic, which judges a sentence to be grammatical if the LM assigns it a higher probability than its local perturbations. We apply this LM-Critic and BIFI along with a large set of unlabeled sentences to bootstrap realistic ungrammatical / grammatical pairs for training a corrector. We evaluate our approach on GEC datasets on multiple domains (CoNLL-2014, BEA-2019, GMEG-wiki and GMEG-yahoo) and show that it outperforms existing methods in both the unsupervised setting (+7.7 F0.5) and the supervised setting (+0.5 F0.5).",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yasunaga-etal-2021-lm">
<titleInfo>
<title>LM-Critic: Language Models for Unsupervised Grammatical Error Correction</title>
</titleInfo>
<name type="personal">
<namePart type="given">Michihiro</namePart>
<namePart type="family">Yasunaga</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jure</namePart>
<namePart type="family">Leskovec</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Percy</namePart>
<namePart type="family">Liang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2021-nov</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Online and Punta Cana, Dominican Republic</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Grammatical error correction (GEC) requires a set of labeled ungrammatical / grammatical sentence pairs for training, but obtaining such annotation can be prohibitively expensive. Recently, the Break-It-Fix-It (BIFI) framework has demonstrated strong results on learning to repair a broken program without any labeled examples, but this relies on a perfect critic (e.g., a compiler) that returns whether an example is valid or not, which does not exist for the GEC task. In this work, we show how to leverage a pretrained language model (LM) in defining an LM-Critic, which judges a sentence to be grammatical if the LM assigns it a higher probability than its local perturbations. We apply this LM-Critic and BIFI along with a large set of unlabeled sentences to bootstrap realistic ungrammatical / grammatical pairs for training a corrector. We evaluate our approach on GEC datasets on multiple domains (CoNLL-2014, BEA-2019, GMEG-wiki and GMEG-yahoo) and show that it outperforms existing methods in both the unsupervised setting (+7.7 F0.5) and the supervised setting (+0.5 F0.5).</abstract>
<identifier type="citekey">yasunaga-etal-2021-lm</identifier>
<identifier type="doi">10.18653/v1/2021.emnlp-main.611</identifier>
<location>
<url>https://aclanthology.org/2021.emnlp-main.611</url>
</location>
<part>
<date>2021-nov</date>
<extent unit="page">
<start>7752</start>
<end>7763</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T LM-Critic: Language Models for Unsupervised Grammatical Error Correction
%A Yasunaga, Michihiro
%A Leskovec, Jure
%A Liang, Percy
%S Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing
%D 2021
%8 nov
%I Association for Computational Linguistics
%C Online and Punta Cana, Dominican Republic
%F yasunaga-etal-2021-lm
%X Grammatical error correction (GEC) requires a set of labeled ungrammatical / grammatical sentence pairs for training, but obtaining such annotation can be prohibitively expensive. Recently, the Break-It-Fix-It (BIFI) framework has demonstrated strong results on learning to repair a broken program without any labeled examples, but this relies on a perfect critic (e.g., a compiler) that returns whether an example is valid or not, which does not exist for the GEC task. In this work, we show how to leverage a pretrained language model (LM) in defining an LM-Critic, which judges a sentence to be grammatical if the LM assigns it a higher probability than its local perturbations. We apply this LM-Critic and BIFI along with a large set of unlabeled sentences to bootstrap realistic ungrammatical / grammatical pairs for training a corrector. We evaluate our approach on GEC datasets on multiple domains (CoNLL-2014, BEA-2019, GMEG-wiki and GMEG-yahoo) and show that it outperforms existing methods in both the unsupervised setting (+7.7 F0.5) and the supervised setting (+0.5 F0.5).
%R 10.18653/v1/2021.emnlp-main.611
%U https://aclanthology.org/2021.emnlp-main.611
%U https://doi.org/10.18653/v1/2021.emnlp-main.611
%P 7752-7763
Markdown (Informal)
[LM-Critic: Language Models for Unsupervised Grammatical Error Correction](https://aclanthology.org/2021.emnlp-main.611) (Yasunaga et al., EMNLP 2021)
ACL