{"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/greedy-search-with-probabilistic-n-gram","title":"Greedy Search with Probabilistic N-gram Matching for Neural Machine Translation","arxiv_id":"1809.03132","date":"2018-09-10","proceeding":"EMNLP 2018 10","authors":["Chenze Shao","Yang Feng","Xilin Chen"],"abstract":"Neural machine translation (NMT) models are usually trained with the\nword-level loss using the teacher forcing algorithm, which not only evaluates\nthe translation improperly but also suffers from exposure bias. Sequence-level\ntraining under the reinforcement framework can mitigate the problems of the\nword-level loss, but its performance is unstable due to the high variance of\nthe gradient estimation. On these grounds, we present a method with a\ndifferentiable sequence-level training objective based on probabilistic n-gram\nmatching which can avoid the reinforcement framework. In addition, this method\nperforms greedy search in the training which uses the predicted words as\ncontext just as at inference to alleviate the problem of exposure bias.\nExperiment results on the NIST Chinese-to-English translation tasks show that\nour method significantly outperforms the reinforcement-based algorithms and\nachieves an improvement of 1.5 BLEU points on average over a strong baseline\nsystem.","url_abs":"http://arxiv.org/abs/1809.03132v1","url_pdf":"http://arxiv.org/pdf/1809.03132v1.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":"greedy-search-with-probabilistic-n-gram","repo_url":"https://github.com/ictnlp/GS4NMT","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"nmt","task_name":"NMT"},{"task_slug":"translation","task_name":"Translation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1809.03132","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}