{"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/word-subword-or-character-an-empirical-study","title":"Word, Subword or Character? An Empirical Study of Granularity in Chinese-English NMT","arxiv_id":"1711.04457","date":"2017-11-13","proceeding":null,"authors":["Yining Wang","Long Zhou","Jiajun Zhang","Cheng-qing Zong"],"abstract":"Neural machine translation (NMT), a new approach to machine translation, has\nbeen proved to outperform conventional statistical machine translation (SMT)\nacross a variety of language pairs. Translation is an open-vocabulary problem,\nbut most existing NMT systems operate with a fixed vocabulary, which causes the\nincapability of translating rare words. This problem can be alleviated by using\ndifferent translation granularities, such as character, subword and hybrid\nword-character. Translation involving Chinese is one of the most difficult\ntasks in machine translation, however, to the best of our knowledge, there has\nnot been any other work exploring which translation granularity is most\nsuitable for Chinese in NMT. In this paper, we conduct an extensive comparison\nusing Chinese-English NMT as a case study. Furthermore, we discuss the\nadvantages and disadvantages of various translation granularities in detail.\nOur experiments show that subword model performs best for Chinese-to-English\ntranslation with the vocabulary which is not so big while hybrid word-character\nmodel is most suitable for English-to-Chinese translation. Moreover,\nexperiments of different granularities show that Hybrid_BPE method can achieve\nbest result on Chinese-to-English translation task.","url_abs":"http://arxiv.org/abs/1711.04457v1","url_pdf":"http://arxiv.org/pdf/1711.04457v1.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":"word-subword-or-character-an-empirical-study","repo_url":"https://github.com/ye-kyaw-thu/myword","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"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":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}