{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/approaching-neural-grammatical-error","title":"Approaching Neural Grammatical Error Correction as a Low-Resource Machine Translation Task","arxiv_id":"1804.05940","date":"2018-04-16","proceeding":"NAACL 2018 6","authors":["Marcin Junczys-Dowmunt","Roman Grundkiewicz","Shubha Guha","Kenneth Heafield"],"abstract":"Previously, neural methods in grammatical error correction (GEC) did not\nreach state-of-the-art results compared to phrase-based statistical machine\ntranslation (SMT) baselines. We demonstrate parallels between neural GEC and\nlow-resource neural MT and successfully adapt several methods from low-resource\nMT to neural GEC. We further establish guidelines for trustable results in\nneural GEC and propose a set of model-independent methods for neural GEC that\ncan be easily applied in most GEC settings. Proposed methods include adding\nsource-side noise, domain-adaptation techniques, a GEC-specific\ntraining-objective, transfer learning with monolingual data, and ensembling of\nindependently trained GEC models and language models. The combined effects of\nthese methods result in better than state-of-the-art neural GEC models that\noutperform previously best neural GEC systems by more than 10% M$^2$ on the\nCoNLL-2014 benchmark and 5.9% on the JFLEG test set. Non-neural\nstate-of-the-art systems are outperformed by more than 2% on the CoNLL-2014\nbenchmark and by 4% on JFLEG.","url_abs":"http://arxiv.org/abs/1804.05940v1","url_pdf":"http://arxiv.org/pdf/1804.05940v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"approaching-neural-grammatical-error","repo_url":"https://github.com/grammatical/neural-naacl2018","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"domain-adaptation","task_name":"Domain Adaptation"},{"task_slug":"grammatical-error-correction","task_name":"Grammatical Error Correction"},{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"},{"task_slug":"translation","task_name":"Translation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/grammatical-error-correction-on-conll-2014","task":"Grammatical Error Correction","dataset":"CoNLL-2014 Shared Task","model":"Transformer","rank_in_archive_order":22,"of":23,"metrics":{"F0.5":"55.8"},"uses_additional_data":false},{"leaderboard":"/sota/grammatical-error-correction-on-jfleg","task":"Grammatical Error Correction","dataset":"JFLEG","model":"Transformer","rank_in_archive_order":5,"of":6,"metrics":{"GLEU":"59.9"},"uses_additional_data":false},{"leaderboard":"/sota/grammatical-error-correction-on-restricted","task":"Grammatical Error Correction","dataset":"Restricted","model":"Transformer","rank_in_archive_order":3,"of":4,"metrics":{"F0.5":"55.8"},"uses_additional_data":false},{"leaderboard":"/sota/grammatical-error-correction-on-_restricted_","task":"Grammatical Error Correction","dataset":"_Restricted_","model":"Transformer","rank_in_archive_order":1,"of":2,"metrics":{"GLEU":"59.9"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1804.05940","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}