{"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/optimal-completion-distillation-for-sequence","title":"Optimal Completion Distillation for Sequence Learning","arxiv_id":"1810.01398","date":"2018-10-02","proceeding":"ICLR 2019 5","authors":["Sara Sabour","William Chan","Mohammad Norouzi"],"abstract":"We present Optimal Completion Distillation (OCD), a training procedure for\noptimizing sequence to sequence models based on edit distance. OCD is\nefficient, has no hyper-parameters of its own, and does not require pretraining\nor joint optimization with conditional log-likelihood. Given a partial sequence\ngenerated by the model, we first identify the set of optimal suffixes that\nminimize the total edit distance, using an efficient dynamic programming\nalgorithm. Then, for each position of the generated sequence, we use a target\ndistribution that puts equal probability on the first token of all the optimal\nsuffixes. OCD achieves the state-of-the-art performance on end-to-end speech\nrecognition, on both Wall Street Journal and Librispeech datasets, achieving\n$9.3\\%$ WER and $4.5\\%$ WER respectively.","url_abs":"http://arxiv.org/abs/1810.01398v2","url_pdf":"http://arxiv.org/pdf/1810.01398v2.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":"optimal-completion-distillation-for-sequence","repo_url":"https://github.com/SaeedNajafi/OCD-Learning","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"optimal-completion-distillation-for-sequence","repo_url":"https://github.com/SaeedNajafi/pytorch-ocd","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":null,"task_name":"Position"},{"task_slug":"speech-recognition","task_name":"Speech Recognition"},{"task_slug":"speech-recognition-1","task_name":"speech-recognition"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1810.01398","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}