{"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/impact-of-corpora-quality-on-neural-machine","title":"Impact of Corpora Quality on Neural Machine Translation","arxiv_id":"1810.08392","date":"2018-10-19","proceeding":null,"authors":["Matīss Rikters"],"abstract":"Large parallel corpora that are automatically obtained from the web,\ndocuments or elsewhere often exhibit many corrupted parts that are bound to\nnegatively affect the quality of the systems and models that learn from these\ncorpora. This paper describes frequent problems found in data and such data\naffects neural machine translation systems, as well as how to identify and deal\nwith them. The solutions are summarised in a set of scripts that remove\nproblematic sentences from input corpora.","url_abs":"http://arxiv.org/abs/1810.08392v1","url_pdf":"http://arxiv.org/pdf/1810.08392v1.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":"impact-of-corpora-quality-on-neural-machine","repo_url":"https://github.com/M4t1ss/parallel-corpora-tools","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"translation","task_name":"Translation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/machine-translation-on-wmt-2017-english","task":"Machine Translation","dataset":"WMT 2017 English-Latvian","model":"Transformer trained on highly filtered data","rank_in_archive_order":1,"of":1,"metrics":{"BLEU":"22.89"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt-2017-latvian","task":"Machine Translation","dataset":"WMT 2017 Latvian-English","model":"Transformer trained on highly filtered data","rank_in_archive_order":1,"of":4,"metrics":{"BLEU":"24.37"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt-2018-english-1","task":"Machine Translation","dataset":"WMT 2018 English-Finnish","model":"Transformer trained on highly filtered data","rank_in_archive_order":1,"of":1,"metrics":{"BLEU":"17.40"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt-2018-finnish","task":"Machine Translation","dataset":"WMT 2018 Finnish-English","model":"Transformer trained on highly filtered data","rank_in_archive_order":2,"of":2,"metrics":{"BLEU":"24.00"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1810.08392","atlas_url":"https://app.syntology.ai/?focus=1810.08392","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}