Research And Implementation On Knowledge Graph Based Potential Relation Mining Technique In Non-Public Sector Of Economy

In order to help the All-China Federation of Industry and Commerce to better supervise and guide the non-public economic market,it is more and more necessary to explore the hidden relationships that may exist among non-public economic entities such as personnel,enterprises,and institutions.This article first constructs a knowledge graph based on the data characteristics of the non-public economic field and subsequent mining requirements,and then uses the knowledge graph to mine the hidden relationships that may exist between entities in multiple dimensions.The main tasks completed in this paper are:(1)A method of constructing knowledge graphs in non-public economic fields based on complex data is proposed.The storage of semi/unstructured data such as text requires the steps of named entity recognition,relationship extraction,entity linking,etc.;in addition,the characteristics of non-public economic data and subsequent mining also pose a challenge to the construction of the knowledge graph.(2)A method of entity recognition and relationship extraction based on Bi-LSTM and other algorithms for non-public economy is proposed.For unstructured information,entities and relationships need to be extracted and stored in the knowledge graph.(3)Proposes the definition of the hidden relationship of the non-public economy.Hidden relationships are divided into two hidden relationships between strongly associated entities and hidden relationships between weak/non-associated entities.(4)A method for mining hidden relationships between strongly associated entities based on path reasoning is proposed.For the entities corresponding to the connected nodes in the knowledge graph,the information on the path between the two is used to summarize and find the relationship between the entities.(5)A method for mining weak/non-associated entity hidden relationships based on graph embedding is proposed.For the entities corresponding to the nodes that are not directly connected in the knowledge graph,use the relevant information around the nodes to calculate the similarity in each dimension between the entities to find the correlation between the two.