Attributed and Predictive Entity Embedding for Fine-Grained Entity Typing in Knowledge Bases

Proceedings of the 27th International Conference on Computational Linguistics, pages 282–292 Santa Fe, New Mexico, USA, August 20-26, 2018. 282 Attributed and Predictive Entity Embedding for Fine-Grained Entity Typing in Knowledge Bases Hailong Jin and Lei Hou and Juanzi Li DCST, Tsinghua University Beijing 100084, China Tiansi Dong B-IT, Bonn University 53113 Bonn, Germany {jinhl15@mails.,houlei@mail.,lijuanzi@}tsinghua.edu.cn dongt@bit.uni-bonn.de Abstract Fine-grained entity typing aims at identifying the semantic type of an entity in KB. Type in- formation is very important in knowledge bases, but is unfortunately incomplete even in some large knowledge bases. Limitations of existing methods are either ignoring the structure and type information in KB or requiring large scale annotated corpus. To address these issues, we propose an attributed and predictive entity embedding method, which can fully utilize various kinds of information comprehensively. Extensive experiments on real DBpedia dataset show that our proposed method significantly outperforms 8 state-of-the-art methods, with 3.4% and 2.9% improvement in Mi-F1 and Ma-F1, respectively. 1 Introduction Knowledge about entities in Knowledge Base (KB) is essential for language understanding. This knowl- edge can be attributional (e.g., canFly, hasAge), relational (e.g., marriedTo, bornIn) or type-based (e.g., isFood, isLocation). Type information, which clusters a group of entities with same properties, is valu- able in KB, because it is the glue that holds our mental world together (Murphy, 2004). Types in KB are usually organized as a hierarchical structure, namely type hierarchy. Entity typing assigns types (e.g., Person, Athlete, BasketballPlayer) to an entity (e.g., Yao Ming) in KB, and is a fundamental task in knowledge base construction. Traditional entity typing focuses on a small set of types, such as Person, Location and Or- ganization (Ratinov and Roth, 2009; Nadeau and Sekine, 2007), while fine-grained entity typ- ing assigns more specific types to an entity, which normally forms a type-path in the type hier- archy in KB (Ren et al., 2016a). As shown in Fig. 1, Yao Ming is associated with a type-path /Thing/Agent/Person/Athlete/BasketballPlayer. Fine-grained types (e.g., Athlete and BasketballPlayer) are more informative than coarse-grained types (e.g., Person), because they provide more specific se- mantic information about the entity (Xu et al., 2016). Characterizing an entity with fine-grained types (type-paths) benefits many real applications, such as knowledge base completion (Dong et al., 2014), entity linking (Ling et al., 2015; Durrett and Klein, 2014), relation extraction (Liu et al., 2014), and question answering (Yahya et al., 2014). Many researches have been carried out on fine-grained entity typing in text, which detects local types for an entity mention in a particular plain text from predefined fine-grained entity types (Ling and Weld, 2012; Yogatama et al., 2015; Ren et al., 2016a; Shimaoka et al., 2017; Xin et al., 2018). However, only a few works aim at fine-grained entity typing in knowledge base, that is, inferring missing types for an entity in KB, which is the focus of this paper. Most KBs are incomplete and lack of type information, in part because the world is always changing. For example, 36.53% entities in DBpedia do not …

Content truncated to 500 words