{"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/segmental-recurrent-neural-networks","title":"Segmental Recurrent Neural Networks","arxiv_id":"1511.06018","date":"2015-11-18","proceeding":null,"authors":["Lingpeng Kong","Chris Dyer","Noah A. Smith"],"abstract":"We introduce segmental recurrent neural networks (SRNNs) which define, given\nan input sequence, a joint probability distribution over segmentations of the\ninput and labelings of the segments. Representations of the input segments\n(i.e., contiguous subsequences of the input) are computed by encoding their\nconstituent tokens using bidirectional recurrent neural nets, and these\n\"segment embeddings\" are used to define compatibility scores with output\nlabels. These local compatibility scores are integrated using a global\nsemi-Markov conditional random field. Both fully supervised training -- in\nwhich segment boundaries and labels are observed -- as well as partially\nsupervised training -- in which segment boundaries are latent -- are\nstraightforward. Experiments on handwriting recognition and joint Chinese word\nsegmentation/POS tagging show that, compared to models that do not explicitly\nrepresent segments such as BIO tagging schemes and connectionist temporal\nclassification (CTC), SRNNs obtain substantially higher accuracies.","url_abs":"http://arxiv.org/abs/1511.06018v2","url_pdf":"http://arxiv.org/pdf/1511.06018v2.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":"segmental-recurrent-neural-networks","repo_url":"https://github.com/ykrmm/ICLR_2020","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"segmental-recurrent-neural-networks","repo_url":"https://github.com/ykrmm/TREMBA","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"chinese-word-segmentation","task_name":"Chinese Word Segmentation"},{"task_slug":"handwriting-recognition","task_name":"Handwriting Recognition"},{"task_slug":"pos","task_name":"POS"},{"task_slug":"pos-tagging","task_name":"POS Tagging"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1511.06018","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}