{"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/explicit-state-tracking-with-semi-supervision","title":"Explicit State Tracking with Semi-Supervision for Neural Dialogue Generation","arxiv_id":"1808.10596","date":"2018-08-31","proceeding":null,"authors":["Xisen Jin","Wenqiang Lei","Zhaochun Ren","Hongshen Chen","Shangsong Liang","Yihong Zhao","Dawei Yin"],"abstract":"The task of dialogue generation aims to automatically provide responses given\nprevious utterances. Tracking dialogue states is an important ingredient in\ndialogue generation for estimating users' intention. However, the\n\\emph{expensive nature of state labeling} and the \\emph{weak interpretability}\nmake the dialogue state tracking a challenging problem for both task-oriented\nand non-task-oriented dialogue generation: For generating responses in\ntask-oriented dialogues, state tracking is usually learned from manually\nannotated corpora, where the human annotation is expensive for training; for\ngenerating responses in non-task-oriented dialogues, most of existing work\nneglects the explicit state tracking due to the unlimited number of dialogue\nstates.\n  In this paper, we propose the \\emph{semi-supervised explicit dialogue state\ntracker} (SEDST) for neural dialogue generation. To this end, our approach has\ntwo core ingredients: \\emph{CopyFlowNet} and \\emph{posterior regularization}.\nSpecifically, we propose an encoder-decoder architecture, named\n\\emph{CopyFlowNet}, to represent an explicit dialogue state with a\nprobabilistic distribution over the vocabulary space. To optimize the training\nprocedure, we apply a posterior regularization strategy to integrate indirect\nsupervision. Extensive experiments conducted on both task-oriented and\nnon-task-oriented dialogue corpora demonstrate the effectiveness of our\nproposed model. Moreover, we find that our proposed semi-supervised dialogue\nstate tracker achieves a comparable performance as state-of-the-art supervised\nlearning baselines in state tracking procedure.","url_abs":"http://arxiv.org/abs/1808.10596v1","url_pdf":"http://arxiv.org/pdf/1808.10596v1.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":"explicit-state-tracking-with-semi-supervision","repo_url":"https://github.com/AuCson/SEDST","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"explicit-state-tracking-with-semi-supervision","repo_url":"https://github.com/shizhediao/SEDST3","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"decoder","task_name":"Decoder"},{"task_slug":"dialogue-generation","task_name":"Dialogue Generation"},{"task_slug":"dialogue-state-tracking","task_name":"Dialogue State Tracking"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1808.10596","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}