Joint Extraction of Entities and Relations Based on a Novel Tagging Scheme

Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1227–1236 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1113 Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1227–1236 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1113 Joint Extraction of Entities and Relations Based on a Novel Tagging Scheme Suncong Zheng, Feng Wang, Hongyun Bao, Yuexing Hao,Peng Zhou, Bo Xu Institute of Automation, Chinese Academy of Sciences, 100190, Beijing, P.R. China {suncong.zheng, feng.wang,hongyun.bao, haoyuexing2014, peng.zhou,xubo}@ia.ac.cn Abstract Joint extraction of entities and relations is an important task in information extrac- tion. To tackle this problem, we firstly propose a novel tagging scheme that can convert the joint extraction task to a tag- ging problem. Then, based on our tag- ging scheme, we study different end-to- end models to extract entities and their re- lations directly, without identifying enti- ties and relations separately. We conduct experiments on a public dataset produced by distant supervision method and the ex- perimental results show that the tagging based methods are better than most of the existing pipelined and joint learning meth- ods. What’s more, the end-to-end model proposed in this paper, achieves the best results on the public dataset. 1 Introduction Joint extraction of entities and relations is to de- tect entity mentions and recognize their semantic relations simultaneously from unstructured text, as Figure 1 shows. Different from open information extraction (Open IE) (Banko et al., 2007) whose relation words are extracted from the given sen- tence, in this task, relation words are extracted from a predefined relation set which may not ap- pear in the given sentence. It is an important issue in knowledge extraction and automatic construc- tion of knowledge base. Traditional methods handle this task in a pipelined manner, i.e., extracting the entities (Nadeau and Sekine, 2007) first and then recog- nizing their relations (Rink, 2010). This separated framework makes the task easy to deal with, and each component can be more flexible. But it ne- glects the relevance between these two sub-tasks The [United States]E-loc President [Trump]E-per will visit the [Apple Inc]E-Org . Country-President None None Extracted Results {United States, Country-President, Trump} Figure 1: A standard example sentence for the task. “Country-President” is a relation in the pre- defined relation set. and each subtask is an independent model. The results of entity recognition may affect the perfor- mance of relation classification and lead to erro- neous delivery (Li and Ji, 2014). Different from the pipelined methods, join- t learning framework is to extract entities togeth- er with relations using a single model. It can ef- fectively integrate the information of entities and relations, and it has been shown to achieve bet- ter results in this task. However, most existing joint methods are feature-based structured system- s (Li and Ji, 2014; Miwa and Sasaki, 2014; Yu and Lam, 2010; Ren et al., 2017). They need complicated feature engineering and heavily re- ly on …

Content truncated to 500 words