{"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/multilingual-relation-extraction-using","title":"Multilingual Relation Extraction using Compositional Universal Schema","arxiv_id":"1511.06396","date":"2015-11-19","proceeding":"NAACL 2016 6","authors":["Patrick Verga","David Belanger","Emma Strubell","Benjamin Roth","Andrew McCallum"],"abstract":"Universal schema builds a knowledge base (KB) of entities and relations by\njointly embedding all relation types from input KBs as well as textual patterns\nexpressing relations from raw text. In most previous applications of universal\nschema, each textual pattern is represented as a single embedding, preventing\ngeneralization to unseen patterns. Recent work employs a neural network to\ncapture patterns' compositional semantics, providing generalization to all\npossible input text. In response, this paper introduces significant further\nimprovements to the coverage and flexibility of universal schema relation\nextraction: predictions for entities unseen in training and multilingual\ntransfer learning to domains with no annotation. We evaluate our model through\nextensive experiments on the English and Spanish TAC KBP benchmark,\noutperforming the top system from TAC 2013 slot-filling using no handwritten\npatterns or additional annotation. We also consider a multilingual setting in\nwhich English training data entities overlap with the seed KB, but Spanish text\ndoes not. Despite having no annotation for Spanish data, we train an accurate\npredictor, with additional improvements obtained by tying word embeddings\nacross languages. Furthermore, we find that multilingual training improves\nEnglish relation extraction accuracy. Our approach is thus suited to\nbroad-coverage automated knowledge base construction in a variety of languages\nand domains.","url_abs":"http://arxiv.org/abs/1511.06396v2","url_pdf":"http://arxiv.org/pdf/1511.06396v2.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":"multilingual-relation-extraction-using","repo_url":"https://github.com/patverga/torch-relation-extraction","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"torch","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"knowledge-base-construction","task_name":"Knowledge Base Construction"},{"task_slug":null,"task_name":"Relation"},{"task_slug":"relation-extraction","task_name":"Relation Extraction"},{"task_slug":"slot-filling","task_name":"Slot Filling"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"},{"task_slug":"word-embeddings","task_name":"Word Embeddings"},{"task_slug":"slot-filling-1","task_name":"slot-filling"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1511.06396","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}