{"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/phrase-based-neural-unsupervised-machine","title":"Phrase-Based & Neural Unsupervised Machine Translation","arxiv_id":"1804.07755","date":"2018-04-20","proceeding":"EMNLP 2018 10","authors":["Guillaume Lample","Myle Ott","Alexis Conneau","Ludovic Denoyer","Marc'Aurelio Ranzato"],"abstract":"Machine translation systems achieve near human-level performance on some\nlanguages, yet their effectiveness strongly relies on the availability of large\namounts of parallel sentences, which hinders their applicability to the\nmajority of language pairs. This work investigates how to learn to translate\nwhen having access to only large monolingual corpora in each language. We\npropose two model variants, a neural and a phrase-based model. Both versions\nleverage a careful initialization of the parameters, the denoising effect of\nlanguage models and automatic generation of parallel data by iterative\nback-translation. These models are significantly better than methods from the\nliterature, while being simpler and having fewer hyper-parameters. On the\nwidely used WMT'14 English-French and WMT'16 German-English benchmarks, our\nmodels respectively obtain 28.1 and 25.2 BLEU points without using a single\nparallel sentence, outperforming the state of the art by more than 11 BLEU\npoints. On low-resource languages like English-Urdu and English-Romanian, our\nmethods achieve even better results than semi-supervised and supervised\napproaches leveraging the paucity of available bitexts. Our code for NMT and\nPBSMT is publicly available.","url_abs":"http://arxiv.org/abs/1804.07755v2","url_pdf":"http://arxiv.org/pdf/1804.07755v2.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":"phrase-based-neural-unsupervised-machine","repo_url":"https://github.com/facebookresearch/UnsupervisedMT","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"phrase-based-neural-unsupervised-machine","repo_url":"https://github.com/1-punchMan/CLTS","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"phrase-based-neural-unsupervised-machine","repo_url":"https://github.com/Helsinki-NLP/shared-info","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"phrase-based-neural-unsupervised-machine","repo_url":"https://github.com/Kelechukwu1/PidginUNMT","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"phrase-based-neural-unsupervised-machine","repo_url":"https://github.com/deterministic-algorithms-lab/Large-XLM","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"phrase-based-neural-unsupervised-machine","repo_url":"https://github.com/facebookresearch/UnsupervisedQA","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"phrase-based-neural-unsupervised-machine","repo_url":"https://github.com/facebookresearch/XLM","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"phrase-based-neural-unsupervised-machine","repo_url":"https://github.com/feyzaakyurek/XLM-LwLL","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"phrase-based-neural-unsupervised-machine","repo_url":"https://github.com/fshdnc/enfi-XLM","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"phrase-based-neural-unsupervised-machine","repo_url":"https://github.com/keleog/PidginUNMT","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"phrase-based-neural-unsupervised-machine","repo_url":"https://github.com/kheeong/XLM_OWN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"phrase-based-neural-unsupervised-machine","repo_url":"https://github.com/luofuli/DualLanST","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"phrase-based-neural-unsupervised-machine","repo_url":"https://github.com/luofuli/DualRL","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"phrase-based-neural-unsupervised-machine","repo_url":"https://github.com/huggingface/transformers","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"nmt","task_name":"NMT"},{"task_slug":"sentence","task_name":"Sentence"},{"task_slug":"translation","task_name":"Translation"},{"task_slug":"unsupervised-machine-translation","task_name":"Unsupervised Machine Translation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/machine-translation-on-wmt2014-english-french","task":"Machine Translation","dataset":"WMT2014 English-French","model":"Unsupervised PBSMT","rank_in_archive_order":52,"of":57,"metrics":{"BLEU score":"28.11"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2014-english-french","task":"Machine Translation","dataset":"WMT2014 English-French","model":"PBSMT + NMT","rank_in_archive_order":53,"of":57,"metrics":{"BLEU score":"27.6"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2014-english-french","task":"Machine Translation","dataset":"WMT2014 English-French","model":"Unsupervised NMT + Transformer","rank_in_archive_order":56,"of":57,"metrics":{"BLEU score":"25.14"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2014-english-german","task":"Machine Translation","dataset":"WMT2014 English-German","model":"PBSMT + NMT","rank_in_archive_order":81,"of":91,"metrics":{"BLEU score":"20.23"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2014-english-german","task":"Machine Translation","dataset":"WMT2014 English-German","model":"Unsupervised PBSMT","rank_in_archive_order":85,"of":91,"metrics":{"BLEU score":"17.94"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2014-english-german","task":"Machine Translation","dataset":"WMT2014 English-German","model":"Unsupervised NMT + Transformer","rank_in_archive_order":87,"of":91,"metrics":{"BLEU score":"17.16"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-english-1","task":"Machine Translation","dataset":"WMT2016 English-Romanian","model":"PBSMT + NMT","rank_in_archive_order":16,"of":21,"metrics":{"BLEU score":"25.13"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-english-1","task":"Machine Translation","dataset":"WMT2016 English-Romanian","model":"Unsupervised PBSMT","rank_in_archive_order":17,"of":21,"metrics":{"BLEU score":"21.33"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-english-1","task":"Machine Translation","dataset":"WMT2016 English-Romanian","model":"Unsupervised NMT + Transformer","rank_in_archive_order":18,"of":21,"metrics":{"BLEU score":"21.18"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-english","task":"Machine Translation","dataset":"WMT2016 English-Russian","model":"PBSMT + NMT","rank_in_archive_order":2,"of":4,"metrics":{"BLEU score":"13.76"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-english","task":"Machine Translation","dataset":"WMT2016 English-Russian","model":"Unsupervised PBSMT","rank_in_archive_order":3,"of":4,"metrics":{"BLEU score":"13.37"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-english","task":"Machine Translation","dataset":"WMT2016 English-Russian","model":"Unsupervised NMT + Transformer","rank_in_archive_order":4,"of":4,"metrics":{"BLEU score":"7.98"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-machine-translation-on-wmt2014-2","task":"Unsupervised Machine Translation","dataset":"WMT2014 English-French","model":"PBSMT + NMT","rank_in_archive_order":7,"of":7,"metrics":{"BLEU":"27.6"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-machine-translation-on-wmt2014-1","task":"Unsupervised Machine Translation","dataset":"WMT2014 French-English","model":"PBSMT + NMT","rank_in_archive_order":6,"of":7,"metrics":{"BLEU":"27.7"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-machine-translation-on-wmt2016","task":"Unsupervised Machine Translation","dataset":"WMT2016 English-German","model":"PBSMT + NMT","rank_in_archive_order":6,"of":7,"metrics":{"BLEU":"20.2"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-machine-translation-on-wmt2016-1","task":"Unsupervised Machine Translation","dataset":"WMT2016 German-English","model":"PBSMT","rank_in_archive_order":7,"of":7,"metrics":{"BLEU":"25.2"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1804.07755","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}