Joint Detection of Topic Entity and Relation for Simple Question Answering

Yunqi Qiu, Yuanzhuo Wang, Xiaolong Jin

Abstract: Knowledge Base is a machine-readable set composed of well-structured relation information between entities, and has become an essential role in automatic question answering. There are two components significant to Knowledge Base Question Answering, i.e., topic entity detection which aims to find out the entity of interest in a given question, and relation detection which aims to find out the relations relevant to the question. Traditional methods decouple these two components, ignoring the correspondence between them. In this paper, we propose a neural attention-based model, namely, Joint Detection Network, to simultaneously detect topic entities and relations for simple question answering. This model can be trained in an end-to-end manner with weak supervision. Experimental results demonstrate the effectiveness of the proposed method.