Neural Relation Extraction with Multi-lingual Attention

Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 34–43 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1004 Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 34–43 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1004 Neural Relation Extraction with Multi-lingual Attention Yankai Lin1, Zhiyuan Liu1∗, Maosong Sun1,2 1 Department of Computer Science and Technology, State Key Lab on Intelligent Technology and Systems, National Lab for Information Science and Technology, Tsinghua University, Beijing, China 2 Jiangsu Collaborative Innovation Center for Language Competence, Jiangsu, China Abstract Relation extraction has been widely used for finding unknown relational facts from the plain text. Most existing methods fo- cus on exploiting mono-lingual data for relation extraction, ignoring massive in- formation from the texts in various lan- guages. To address this issue, we intro- duce a multi-lingual neural relation ex- traction framework, which employs mono- lingual attention to utilize the information within mono-lingual texts and further pro- poses cross-lingual attention to consider the information consistency and comple- mentarity among cross-lingual texts. Ex- perimental results on real-world datasets show that our model can take advan- tage of multi-lingual texts and consistently achieve significant improvements on re- lation extraction as compared with base- lines. The source code of this paper can be obtained from https://github. com/thunlp/MNRE 1 Introduction People build many large-scale knowledge bases (KBs) to store structured knowledge about the real world, such as Wikidata1 and DBpedia2. KBs are playing an important role in many AI and NLP applications such as information retrieval and question answering. The facts in KBs are typically organized in the form of triplets, e.g., (New York, CityOf, United States). Since ex- isting KBs are far from complete and new facts are growing infinitely, meanwhile manual anno- tation of these knowledge is time-consuming and ∗ Corresponding author: Zhiyuan Liu (li- uzy@tsinghua.edu.cn). 1http://www.wikidata.org/ 2http://wiki.dbpedia.org/ human-intensive, many works have been devoted to automated extraction of novel facts from vari- ous Web resources, where relation extraction (RE) from plain texts is one the most important knowl- edge sources. Among various methods for relation extraction, distant supervision is the most promising approach (Mintz et al., 2009; Riedel et al., 2010; Hoffmann et al., 2011; Surdeanu et al., 2012), which can au- tomatically generate training instances via aligning KBs and texts to address the issue of lacking super- vised data. As the development of deep learning, Zeng et al. (2015) introduce neural networks to ex- tract relations with automatically learned features from training instances. To address the wrong labelling issue of distant supervision data, Lin et al. (2016) further employ sentence-level atten- tion mechanism in neural relation extraction, and achieves the state-of-the-art performance. However, most RE systems concentrate on ex- tracting relational facts from mono-lingual data. In fact, people describe knowledge about the world using various languages. And people speaking different languages also share similar knowledge about the world due to the similarities of human experiences and …

Content truncated to 500 words