Adaptive Attentional Network for Few-Shot Knowledge Graph Completion

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 1681–1691, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 1681 Adaptive Attentional Network for Few-Shot Knowledge Graph Completion Jiawei Sheng1,2, Shu Guo3, Zhenyu Chen4, Juwei Yue4, Lihong Wang3∗, Tingwen Liu1,2 and Hongbo Xu1,2 1 Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China 2 School of Cyber Security, University of Chinese Academy of Sciences, Beijing, China 3 National Computer Network Emergency Response Technical Team/Coordination Center of China 4 Beijing Advanced Innovation Center of Big Data and Brain Computing, Beihang University shengjiawei@iie.ac.cn, guoshu@cert.org.cn, wlh@isc.org.cn Abstract Few-shot Knowledge Graph (KG) completion is a focus of current research, where each task aims at querying unseen facts of a rela- tion given its few-shot reference entity pairs. Recent attempts solve this problem by learn- ing static representations of entities and refer- ences, ignoring their dynamic properties, i.e., entities may exhibit diverse roles within task relations, and references may make different contributions to queries. This work proposes an adaptive attentional network for few-shot KG completion by learning adaptive entity and reference representations. Specifically, en- tities are modeled by an adaptive neighbor encoder to discern their task-oriented roles, while references are modeled by an adaptive query-aware aggregator to differentiate their contributions. Through the attention mecha- nism, both entities and references can capture their fine-grained semantic meanings, and thus render more expressive representations. This will be more predictive for knowledge acqui- sition in the few-shot scenario. Evaluation in link prediction on two public datasets shows that our approach achieves new state-of-the- art results with different few-shot sizes. The source code is available at https://github. com/JiaweiSheng/FAAN. 1 Introduction Knowledge Graphs (KGs) like Freebase (Bollacker et al., 2008), NELL (Carlson et al., 2010) and Wiki- data (Vrandecic and Kr¨otzsch, 2014) are extremely useful resources for NLP tasks, such as information extraction (Liu et al., 2018), machine reading (Yang and Mitchell, 2017), and relation extraction (Ren et al., 2017). A typical KG is a multi-relational graph, represented as triples of the form (h, r, t), indicating that two entities are connected by rela- tion r. Although a KG contains a great number ∗Corresponding author: Lihong Wang. Bill Gates Jennifer Gates Melinda Gates Microsoft Chairman Paul Allen WorkWith ProxyFor HasJobPosition MarryTo HasChild HasChild Rory Gates CeoOf ? ParentOfPerson ? Reference Query (Petersburg, SubPartOf, Virginia) (Vacaville, SubPartOf, California) (Prague, SubPartOf, Czech) (Cavaliers, SubPartOf, NBA) (Los Angeles Lakers, SubPartOf, NBA) (Chicago Bulls, SubPartOf, NBA) (b) (a) Figure 1: Illustration of dynamic properties in few-shot KG completion: (a) An entity has diverse roles in dif- ferent tasks; and (b) References show distinct contribu- tions to a particular query. of triples, it is also known to suffer from incom- pleteness problem. KG completion, which aims at automatically inferring missing facts by examining existing ones, has thus attracted broad attention. A promising approach, namely KG embedding, has been proposed and successfully applied to this task. The key idea is to embed KG components, includ- ing entities and relations, into a continuous vector …

Content truncated to 500 words