{"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/linguistic-input-features-improve-neural","title":"Linguistic Input Features Improve Neural Machine Translation","arxiv_id":"1606.02892","date":"2016-06-09","proceeding":"WS 2016 8","authors":["Rico Sennrich","Barry Haddow"],"abstract":"Neural machine translation has recently achieved impressive results, while\nusing little in the way of external linguistic information. In this paper we\nshow that the strong learning capability of neural MT models does not make\nlinguistic features redundant; they can be easily incorporated to provide\nfurther improvements in performance. We generalize the embedding layer of the\nencoder in the attentional encoder--decoder architecture to support the\ninclusion of arbitrary features, in addition to the baseline word feature. We\nadd morphological features, part-of-speech tags, and syntactic dependency\nlabels as input features to English<->German, and English->Romanian neural\nmachine translation systems. In experiments on WMT16 training and test sets, we\nfind that linguistic input features improve model quality according to three\nmetrics: perplexity, BLEU and CHRF3. An open-source implementation of our\nneural MT system is available, as are sample files and configurations.","url_abs":"http://arxiv.org/abs/1606.02892v2","url_pdf":"http://arxiv.org/pdf/1606.02892v2.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":"linguistic-input-features-improve-neural","repo_url":"https://github.com/rsennrich/wmt16-scripts","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"decoder","task_name":"Decoder"},{"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-wmt2016-english-german","task":"Machine Translation","dataset":"WMT2016 English-German","model":"Linguistic Input Features","rank_in_archive_order":3,"of":12,"metrics":{"BLEU score":"28.4"},"uses_additional_data":false},{"leaderboard":"/sota/machine-translation-on-wmt2016-german-english","task":"Machine Translation","dataset":"WMT2016 German-English","model":"Linguistic Input Features","rank_in_archive_order":4,"of":8,"metrics":{"BLEU score":"32.9"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1606.02892","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}