面向多元时序关系的金融知识图谱表示与构建  

Hypergraph Based Knowledge Representation and Construction for Polynary and Temporal Relations in Financial Domain

作　　者：高峰 郑丽丽[1,2,3,4] 顾进广 GAO Feng;ZHENG Lili;GU Jin’guang(Department of Computer Science and Technology,Wuhan University of Science and Technology,Wuhan 430065,China;Big Data Science and Engineering Research Institute,Wuhan University of Science and Technology,Wuhan 430065,China;Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial,Wuhan 430065,China;Key Laboratory of Content Organization and Knowledge Service for Rich Media Digital Publishing,Wuhan 430065,China)

机构地区：[1]武汉科技大学计算机科学与技术学院,湖北武汉430065 [2]武汉科技大学大数据科学与工程研究院,湖北武汉430065 [3]湖北省智能信息处理与实时工业系统重点实验室,湖北武汉430065 [4]富媒体数字出版内容组织与知识服务重点实验室,湖北武汉430065

出　　处：《山西大学学报（自然科学版）》2022年第4期873-883,共11页Journal of Shanxi University(Natural Science Edition)

基　　金：科技创新2030“新一代人工智能”重大项目(2020AAA0108500);国家自然科学基金(U1836118);富媒体数字出版内容组织与知识服务重点实验室开放基金(ZD2021-11/01)。

摘　　要：对共同担保、交叉持股、关联交易等多元关系和事件时序进行表示是金融领域知识表示的重要需求,但目前的知识图谱表示机制以二元关系为基础,且没有建立恰当的金融知识时序表示和推理机制。为解决以上两个问题,文章以OWL(Web Ontology Language)语义框架为基础,研发了一套针对金融领域知识的时序超图表示模型。同时,在保证对现有金融知识的兼容性基础上,通过定义金融领域时序和多元关系的推理规则并扩展现有推理机,实现了多元时序金融超图的自动构建。实验表明,金融时序超图对于表示金融知识和事件具有通用性和灵活性,且本文模型检索时间仅为传统二元模型的5%左右。The representation of temporal relation and polynary relations,such as together guarantee,cross-shareholdings,relatedparty transactions,is one of important requirements for knowledge representation in the financial field.However,the current knowledge graph representation mechanism is based on dual relationships and does not establish the appropriate temporal representation and reasoning mechanism of financial knowledge.In order to solve the above two problems,the financial temporal hypergraph based on the OWL(Web Ontology Language)semantic framework was studied,and on the basis of ensuring the compatibility of existing financial knowledge,this paper realizes the automatic construction of financial temporal hypergraph by defining the reasoning rules and expanding the existing reasoning machine.The experiments show that the financial graph has universality and flexibility for representing financial knowledge and events,and the retrieval time of the model is about 5%of the traditional binary representation model.

关 键 词：多元关系 事件时序 金融知识图谱 超图 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]