Learning to Link Entities with Knowledge Base

Human Language Technologies: The 2010 Annual Conference of the North American Chapter of the ACL, pages 483–491, Los Angeles, California, June 2010. c⃝2010 Association for Computational Linguistics Learning to Link Entities with Knowledge Base Zhicheng Zheng, Fangtao Li, Minlie Huang, Xiaoyan Zhu State Key Laboratory of Intelligent Technology and Systems Tsinghua National Laboratory for Information Science and Technology Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China { zhengzc04,fangtao06} @gmail.com, { aihuang,zxy-dcs} @tsinghua.edu.cn Abstract This paper address the problem of entity link- ing. Specifically, given an entity mentioned in unstructured texts, the task is to link this entity with an entry stored in the existing knowledge base. This is an important task for informa- tion extraction. It can serve as a convenient gateway to encyclopedic information, and can greatly improve the web users’ experience. Previous learning based solutions mainly fo- cus on classification framework. However, it’s more suitable to consider it as a ranking prob- lem. In this paper, we propose a learning to rank algorithm for entity linking. It effectively utilizes the relationship information among the candidates when ranking. The experi- ment results on the TAC 20091 dataset demon- strate the effectiveness of our proposed frame- work. The proposed method achieves 18.5% improvement in terms of accuracy over the classification models for those entities which have corresponding entries in the Knowledge Base. The overall performance of the system is also better than that of the state-of-the-art methods. 1 Introduction The entity linking task is to map a named-entity mentioned in a text to a corresponding entry stored in the existing Knowledge Base. The Knowledge Base can be considered as an encyclopedia for en- tities. It contains definitional, descriptive or rele- vant information for each entity. We can acquire the knowledge of entities by looking up the Knowledge 1http://www.nist.gov/tac/ Base. Wikipedia is an online encyclopedia, and now it becomes one of the largest repositories of encyclo- pedic knowledge. In this paper, we use Wikipedia as our Knowledge Base. Entity linking can be used to automatically aug- ment text with links, which serve as a conve- nient gateway to encyclopedic information, and can greatly improve user experience. For example, Fig- ure 1 shows news from BBC.com. When a user is interested in ”Thierry Henry”, he can acquire more detailed information by linking ”Thierry Henry” to the corresponding entry in the Knowledge Base. Figure 1: Entity linking example Entity linking is also useful for some information extraction (IE) applications. We can make use of information stored in the Knowledge Base to assist the IE problems. For example, to answer the ques- tion ”When was the famous basketball player Jor- dan born?”, if the Knowledge Base contains the en- 483 tity of basketball player Michael Jordan and his in- formation (such as infobox2 in Wikipedia), the cor- rect answer ”February 17, 1963” can be easily re- trieved. Entity linking encounters the problem of entity ambiguity. One entity may refer to several entries in the Knowledge Base. For example, the entity ”Michael …

Content truncated to 500 words