{"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/attaining-the-unattainable-reassessing-claims","title":"Attaining the Unattainable? Reassessing Claims of Human Parity in Neural Machine Translation","arxiv_id":"1808.10432","date":"2018-08-30","proceeding":"WS 2018 10","authors":["Antonio Toral","Sheila Castilho","Ke Hu","Andy Way"],"abstract":"We reassess a recent study (Hassan et al., 2018) that claimed that machine\ntranslation (MT) has reached human parity for the translation of news from\nChinese into English, using pairwise ranking and considering three variables\nthat were not taken into account in that previous study: the language in which\nthe source side of the test set was originally written, the translation\nproficiency of the evaluators, and the provision of inter-sentential context.\nIf we consider only original source text (i.e. not translated from another\nlanguage, or translationese), then we find evidence showing that human parity\nhas not been achieved. We compare the judgments of professional translators\nagainst those of non-experts and discover that those of the experts result in\nhigher inter-annotator agreement and better discrimination between human and\nmachine translations. In addition, we analyse the human translations of the\ntest set and identify important translation issues. Finally, based on these\nfindings, we provide a set of recommendations for future human evaluations of\nMT.","url_abs":"http://arxiv.org/abs/1808.10432v1","url_pdf":"http://arxiv.org/pdf/1808.10432v1.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":"attaining-the-unattainable-reassessing-claims","repo_url":"https://github.com/antot/human_parity_mt","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok","spdx":"GPL-3.0"}}],"tasks":[{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"translation","task_name":"Translation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1808.10432","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}