{"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/discofuse-a-large-scale-dataset-for-discourse","title":"DiscoFuse: A Large-Scale Dataset for Discourse-Based Sentence Fusion","arxiv_id":"1902.10526","date":"2019-02-27","proceeding":"NAACL 2019 6","authors":["Mor Geva","Eric Malmi","Idan Szpektor","Jonathan Berant"],"abstract":"Sentence fusion is the task of joining several independent sentences into a\nsingle coherent text. Current datasets for sentence fusion are small and\ninsufficient for training modern neural models. In this paper, we propose a\nmethod for automatically-generating fusion examples from raw text and present\nDiscoFuse, a large scale dataset for discourse-based sentence fusion. We author\na set of rules for identifying a diverse set of discourse phenomena in raw\ntext, and decomposing the text into two independent sentences. We apply our\napproach on two document collections: Wikipedia and Sports articles, yielding\n60 million fusion examples annotated with discourse information required to\nreconstruct the fused text. We develop a sequence-to-sequence model on\nDiscoFuse and thoroughly analyze its strengths and weaknesses with respect to\nthe various discourse phenomena, using both automatic as well as human\nevaluation. Finally, we conduct transfer learning experiments with WebSplit, a\nrecent dataset for text simplification. We show that pretraining on DiscoFuse\nsubstantially improves performance on WebSplit when viewed as a sentence fusion\ntask.","url_abs":"http://arxiv.org/abs/1902.10526v3","url_pdf":"http://arxiv.org/pdf/1902.10526v3.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":"discofuse-a-large-scale-dataset-for-discourse","repo_url":"https://github.com/google-research-datasets/discofuse","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"discofuse-a-large-scale-dataset-for-discourse","repo_url":"https://github.com/hyeonsuukang/augmenting_tochi22","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}}],"tasks":[{"task_slug":"articles","task_name":"Articles"},{"task_slug":"sentence","task_name":"Sentence"},{"task_slug":"sentence-fusion","task_name":"Sentence Fusion"},{"task_slug":"text-simplification","task_name":"Text Simplification"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"}],"methods":[],"datasets_introduced":[{"slug":"discofuse","name":"DiscoFuse","full_name":"DiscoFuse"}],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1902.10526","atlas_url":"https://app.syntology.ai/?focus=1902.10526","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}