Research On Entity Alignment Method Of Knowledge Graph In Domain Of Social Insurances And Housing Fund

Social insurances and housing fund is the general term of several kinds of insurances provided by the state for employees,including endowment insurance,medical insurance,unemployment insurance,work-related injury insurance,and housing accumulation fund.It is the fundamental rights and interests of workers stipulated by the state and an essential guarantee for safeguarding people's living standards and legitimate rights and interests.Social insurances and housing fund policy has played an essential role in establishing and improving China's society.In the constant improvement of the social security system,the state has also issued laws and regulations.Mining the internal relationship of policies and regulations and establishing a social security knowledge system will help the research and audit work in social insurances and housing fund.We have extracted triples from social insurances and housing fund policies by named entity recognition and relationship extraction technology.These triples are mainly from the policies and regulations of Social insurances and housing fund.They are complementary to the knowledge in the encyclopedia knowledge graph constructed by crowdsourcing.Therefore,this paper aims to fusion knowledge graphs by aligning entities in the domain knowledge graphs with the Chinese encyclopedia knowledge graph.The entities with the equivalent relationship are found to realize knowledge fusion by entity links formed between knowledge graphs to expand the knowledge graph of social insurances and housing fund.This paper studies the entity alignment method and application of knowledge graph in social insurance and housing fund based on the above reasons.This dissertation is divided into three parts.The first part is screening candidates and constructing the data set,that is,the screening of entities in the Chinese encyclopedia knowledge graph.Due to the vast number of triples in the Chinese encyclopedia knowledge graph,this paper constructs a candidate entity set by calculating the character similarity of entity names by local sensitive hash technology.To build entity alignment corpus set by manually marking equivalent entities for subsequent research.The second part is the research of the entity alignment algorithm.Existing methods set the same margin for all triples as optimization objectives.Considering that different relationships have different semantic information and scope,this paper thinks that different margins should be set for triples under different relationships.The knowledge representation learning entity alignment algorithm with adaptive margin and attribute triple preprocessing is given.And the entity alignment algorithm by calculating entity description information's text similarity through multi-encoder and feature combination.Finally,the above algorithms' corresponding experiments are carried out on the constructed data set,and these knowledge graphs are combined through the equivalence relationship between entities.The third part is the application of knowledge graph and entity alignment technology in the field of social insurances and housing fund.This paper develops a knowledge search tool based on the knowledge graph.And the visualization of entities and the joint retrieval of policies and regulations in the knowledge graph are realized to show the relevant knowledge in the field of social insurances and housing fund more intuitively.