知识图谱问答领域综述  

Overview on Knowledge Graph Question Answering

作　　者：郑泳智 朱定局[1] 吴惠粦 彭小荣 ZHENG Yong-Zhi;ZHU Ding-Ju;WU Hui-Lin;PENG Xiao-Rong(School of Computer Science,South China Normal University,Guangzhou 510631,China;Guangzhou National Modern Agricultural Industry Science and Technology Innovation Center,Guangzhou 510520,China;Zengcheng District Cultural Center of Guangzhou,Foshan 511300,China)

机构地区：[1]华南师范大学计算机学院,广州510631 [2]广州国家现代农业产业科技创新中心,广州510520 [3]广州市增城区文化馆,佛山511300

出　　处：《计算机系统应用》2022年第4期1-13,共13页Computer Systems & Applications

基　　金：中国高等教育学会专项课题(2020JXD01);广东省普通高校“人工智能”重点领域专项(2019KZDZX1027);广东高校省级重点平台和重大科研项目(2017KTSCX048);广东省公益研究与能力建设(2018B070714018);广东省中医药局科研项目(20191411);广东省高等学校产业学院建设项目(人工智能机器人教育产业学院)。

摘　　要：近年来,随着知识图谱的发展,利用给定的知识图谱数据自动得出人类自然语言问题的答案成为了时下的研究热点,诸如Siri和小爱同学的问答系统已经广泛投入使用.得益于深度学习的引入,该领域的各子课题虽然有所突破,但依然存在需要攻克的难点,例如多跳推理和策略组合等.本文从主流的构建方法为切入点,归纳总结该领域研究现状以及所面临的挑战,不仅有助于研究者高效展开对该领域的研究工作,更有利于不同行业的研究者研发行业相关的问答系统,提高行业生产力.With the development of knowledge graphs,utilizing given knowledge graph data to automatically obtain answers to human natural language questions has become popular in recent years.QA(question answering)systems such as Siri and Xiao Ai have been widely used.Thanks to the introduction of deep learning,breakthroughs have been made in various sub-projects in this field,while there are still difficulties that need to be overcome,such as multi-hop reasoning and strategy combination.Therefore,starting from the mainstream construction method,this study summarizes the current research status and challenges in this field,which can not only help researchers to efficiently carry out research in this field but also help researchers in different industries to develop industry-related QA systems to improve productivity.

关 键 词：知识图谱 智能问答 语义解析 信息检索 问答系统 分析工具 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]