{"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/confidence-through-attention","title":"Confidence through Attention","arxiv_id":"1710.03743","date":"2017-10-10","proceeding":"MTSummit 2017 9","authors":["Matīss Rikters","Mark Fishel"],"abstract":"Attention distributions of the generated translations are a useful bi-product\nof attention-based recurrent neural network translation models and can be\ntreated as soft alignments between the input and output tokens. In this work,\nwe use attention distributions as a confidence metric for output translations.\nWe present two strategies of using the attention distributions: filtering out\nbad translations from a large back-translated corpus, and selecting the best\ntranslation in a hybrid setup of two different translation systems. While\nmanual evaluation indicated only a weak correlation between our confidence\nscore and human judgments, the use-cases showed improvements of up to 2.22 BLEU\npoints for filtering and 0.99 points for hybrid translation, tested on\nEnglish<->German and English<->Latvian translation.","url_abs":"http://arxiv.org/abs/1710.03743v1","url_pdf":"http://arxiv.org/pdf/1710.03743v1.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":"confidence-through-attention","repo_url":"https://github.com/M4t1ss/ConfidenceThroughAttention","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null},{"paper_slug":"confidence-through-attention","repo_url":"https://github.com/CSTR-Edinburgh/ophelia","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"confidence-through-attention","repo_url":"https://github.com/oliverwatts/ophelia","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"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-wmt-2017-latvian","task":"Machine Translation","dataset":"WMT 2017 Latvian-English","model":"Attention-based Hybrid NMT combination","rank_in_archive_order":3,"of":4,"metrics":{"BLEU":"14.83"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1710.03743","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}