基于表示学习的知识库问答研究进展与展望

Representation Learning for Question Answering over Knowledge Base:An Overview

面向知识库的问答(Question answering over knowledge base,KBQA)是问答系统的重要组成.近些年,随着以深度学习为代表的表示学习技术在多个领域的成功应用,许多研究者开始着手研究基于表示学习的知识库问答技术.其基本假设是把知识库问答看做是一个语义匹配的过程.通过表示学习知识库以及用户问题的语义表示,将知识库中的实体、关系以及问句文本转换为一个低维语义空间中的数值向量,在此基础上,利用数值计算,直接匹配与用户问句语义最相似的答案.从目前的结果看,基于表示学习的知识库问答系统在性能上已经超过传统知识库问答方法.本文将对现有基于表示学习的知识库问答的研究进展进行综述,包括知识库表示学习和问句(文本)表示学习的代表性工作,同时对于其中存在难点以及仍存在的研究问题进行分析和讨论.

Question answering over knowledge base（KBQA） is an important direction for the research of question answering. Recently, with the drastic development of deep learning, researchers and developers have paid more attentions to KBQA from this angle. They regarded this problem as a task of semantic matching. The semantics of knowledge base and users questions are learned through representation learning under the framework of deep learning. The entities and relations in knowledge base and the texts in questions could be represented as numerical vectors. Then, the answer could be figured out through similarity computation between the vectors of knowledge base and the vectors of the given question. From reported results, KBQA based on representation learning has obtained the best performance. This paper introduces the mainstream methods in this area. It further induces the typical approaches of representation learning on knowledge base and texts（questions）, respectively. Finally, the current research challenges are discussed.