知识图谱多跳问答推理研究进展、挑战与展望  

Progress,challenges and research trends of reasoning in multi-hop knowledge graph based question answering

作　　者：杜会芳 王昊奋 史英慧 王萌[3] DU Huifang;WANG Haofen;SHI Yinghui;WANG Meng(College of Design and Innovation,Tongji University,Shanghai 200092,China;School of Cyber Science and Engineering,Southeast University,Wuxi 214100,China;School of Computer Science and Engineering,Southeast University,Nanjing 211189,China)

机构地区：[1]同济大学设计创意学院,上海200092 [2]东南大学网络空间与安全学院,江苏无锡214100 [3]东南大学计算机科学与工程学院,江苏南京211189

出　　处：《大数据》2021年第3期60-79,共20页Big Data Research

基　　金：中央高校基本科研业务专项资金资助项目(No.22120210109)。

摘　　要：近年来,知识图谱问答在医疗、金融、政务等领域被广泛应用。用户不再满足于关于实体属性的单跳问答,而是更多地倾向表达复杂的多跳问答需求。为了应对上述复杂多跳问答,各种不同类型的推理方法被陆续提出。系统地介绍了基于嵌入、路径、逻辑的多跳知识问答推理的最新研究进展以及相关数据集和评测指标,并重点围绕前沿问题进行了讨论。最后总结了现有方法的不足,并展望了未来的研究方向。Recently,knowledge graph based question answering has been widely used in many fields such as medical care,finance,and government affairs.Users are no longer satisfied with question answering service of single-hop entity attributes,but want service which can handle complex multi-hop question.In order to accurately and deeply understand multi-hop questions,various types of reasoning methods have been proposed.The latest research methods of multi-hop knowledge graph based question answering were systematically introduced,as well as related datasets and evaluation metrics.These researches methods were divided into three categories:embedding-based methods,path-based methods,and logic-based methods.In particular,frontier issues discussion were focused.Finally,the shortcomings of existing methods and suggest future research trends were summarized.

关 键 词：知识图谱 多跳问答 推理 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]