| 摘 要: 知识图谱是实现对话机器人的一类重要工具。如何通过一套完整流程来构建基于知识图谱的问答系统是比较复杂的。因此，本文从构建基于知识图谱的问答系统的全流程角度总结了多个主题：知识图谱类型、知识图谱构建与存储、应用在知识图谱对话中的语言模型、图空间内的语义匹配及生成。进一步，本文在各主题的垂直领域归纳了常用方法及模型，并分析了各子模块的目的和必要性。最后，本文通过总结出的必要模块及流程，给出了一种基于知识图谱的问答系统的基线模型快速构建方法。该方法借助了各模块的前沿算法且有效地保证了拓展性、准确性和时效性。 |
| 关键词: 知识图谱；问答系统；对话机器人；语言模型；语义匹配 |
|
中图分类号: TP183
文献标识码: A
|
|
| Implementation of Question Answering based on Knowledge Graph |
|
WEI Zelin1,2, ZHANG Shuai1,2, WANG Jianchao1,2
|
( 1. Dalian Neusoft University of Information, Dalian 116023, China ;
2. Research Institute, Dalian Neusoft Education Technology Group Co. Limited, Dalian 116023, China)
weizelin@neuedu.com; zhangshuai@neuedu.com; wangjianchao@neuedu.com
|
| Abstract: Knowledge graph is an important tool for realizing chatbots. The lifecycle of constructing a question answering system based on knowledge graph is a complex task. This paper summarizes a number of topics from the perspective of building a knowledge graph-based question answering system. The topics include knowledge graph types, knowledge graph construction and storage, language models used in knowledge graph dialogue, semantic matching and generation in graph space. Furthermore, this paper summarizes commonly used methods and models in vertical areas of topics, and analyzes the purpose and necessity of sub-modules. A method for quickly constructing a baseline model of a knowledge graph based question answering system will be presented. The proposed method relies on the cutting-edge algorithms and effectively guarantees scalability, accuracy and timeliness. |
| Keywords: knowledge graph; question answering system; chatbot; language model; semantic matching |
