{"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/rst-parsing-from-scratch","title":"RST Parsing from Scratch","arxiv_id":"2105.10861","date":"2021-05-23","proceeding":"NAACL 2021 4","authors":["Thanh-Tung Nguyen","Xuan-Phi Nguyen","Shafiq Joty","XiaoLi Li"],"abstract":"We introduce a novel top-down end-to-end formulation of document-level discourse parsing in the Rhetorical Structure Theory (RST) framework. In this formulation, we consider discourse parsing as a sequence of splitting decisions at token boundaries and use a seq2seq network to model the splitting decisions. Our framework facilitates discourse parsing from scratch without requiring discourse segmentation as a prerequisite; rather, it yields segmentation as part of the parsing process. Our unified parsing model adopts a beam search to decode the best tree structure by searching through a space of high-scoring trees. With extensive experiments on the standard English RST discourse treebank, we demonstrate that our parser outperforms existing methods by a good margin in both end-to-end parsing and parsing with gold segmentation. More importantly, it does so without using any handcrafted features, making it faster and easily adaptable to new languages and domains.","url_abs":"https://arxiv.org/abs/2105.10861v1","url_pdf":"https://arxiv.org/pdf/2105.10861v1.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":"rst-parsing-from-scratch","repo_url":"https://github.com/tungngthanh/rst_parser","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"discourse-parsing","task_name":"Discourse Parsing"},{"task_slug":"discourse-segmentation","task_name":"Discourse Segmentation"},{"task_slug":"end-to-end-rst-parsing","task_name":"End-to-End RST Parsing"},{"task_slug":"segmentation","task_name":"Segmentation"}],"methods":[{"method_slug":"lstm","method_name":"LSTM"},{"method_slug":"seq2seq","method_name":"Seq2Seq"},{"method_slug":"sigmoid-activation","method_name":"Sigmoid Activation"},{"method_slug":"tanh-activation","method_name":"Tanh Activation"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/discourse-parsing-on-rst-dt","task":"Discourse Parsing","dataset":"RST-DT","model":"End-to-end Top-down (XLNet)","rank_in_archive_order":17,"of":40,"metrics":{"RST-Parseval (Nuclearity)":"76.0","RST-Parseval (Relation)":"61.8","RST-Parseval (Span)":"87.6","Standard Parseval (Full)":"50.2","Standard Parseval (Nuclearity)":"64.3","Standard Parseval (Relation)":"51.6","Standard Parseval (Span)":"74.3"},"uses_additional_data":false},{"leaderboard":"/sota/discourse-parsing-on-rst-dt","task":"Discourse Parsing","dataset":"RST-DT","model":"End-to-end Top-down (Glove)","rank_in_archive_order":21,"of":40,"metrics":{"Standard Parseval (Full)":"46.8","Standard Parseval (Nuclearity)":"59.6","Standard Parseval (Relation)":"47.7","Standard Parseval (Span)":"71.1"},"uses_additional_data":false},{"leaderboard":"/sota/end-to-end-rst-parsing-on-rst-dt-1","task":"End-to-End RST Parsing","dataset":"RST-DT","model":"Nguyen et al. (2021)","rank_in_archive_order":4,"of":4,"metrics":{"Standard Parseval (Full)":"46.6","Standard Parseval (Nuclearity)":"59.1","Standard Parseval (Relation)":"47.8","Standard Parseval (Span)":"68.4"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2105.10861","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}