{"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/simultaneously-self-attending-to-all-mentions","title":"Simultaneously Self-Attending to All Mentions for Full-Abstract Biological Relation Extraction","arxiv_id":"1802.10569","date":"2018-02-28","proceeding":"NAACL 2018 6","authors":["Patrick Verga","Emma Strubell","Andrew McCallum"],"abstract":"Most work in relation extraction forms a prediction by looking at a short\nspan of text within a single sentence containing a single entity pair mention.\nThis approach often does not consider interactions across mentions, requires\nredundant computation for each mention pair, and ignores relationships\nexpressed across sentence boundaries. These problems are exacerbated by the\ndocument- (rather than sentence-) level annotation common in biological text.\nIn response, we propose a model which simultaneously predicts relationships\nbetween all mention pairs in a document. We form pairwise predictions over\nentire paper abstracts using an efficient self-attention encoder. All-pairs\nmention scores allow us to perform multi-instance learning by aggregating over\nmentions to form entity pair representations. We further adapt to settings\nwithout mention-level annotation by jointly training to predict named entities\nand adding a corpus of weakly labeled data. In experiments on two Biocreative\nbenchmark datasets, we achieve state of the art performance on the Biocreative\nV Chemical Disease Relation dataset for models without external KB resources.\nWe also introduce a new dataset an order of magnitude larger than existing\nhuman-annotated biological information extraction datasets and more accurate\nthan distantly supervised alternatives.","url_abs":"http://arxiv.org/abs/1802.10569v1","url_pdf":"http://arxiv.org/pdf/1802.10569v1.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":"simultaneously-self-attending-to-all-mentions","repo_url":"https://github.com/patverga/bran","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"all","task_name":"All"},{"task_slug":null,"task_name":"Relation"},{"task_slug":"relation-extraction","task_name":"Relation Extraction"},{"task_slug":"sentence","task_name":"Sentence"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1802.10569","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}