Research On The Construction And Application Of Social Insurance Knowledge Graph Based On Entity And Relation Extraction

As an important component of the Social Security System,Social Insurance is closely related to everyone’s life.With the development of the Internet,people pay more and more attention to Social Insurance.How to quickly obtain the Social Insurance knowledge in the multifarious data ocean has become an urgent problem to be solved.In recent research,Knowledge Graph have received considerable attention due to their ability to intuitively construct structured knowledge systems.Entity and Relation Extraction is capable of extracting structured knowledge representations from unstructured text,and therefore plays an important role in the construction of Knowledge Graph.Therefore,constructing a Knowledge Graph in the Social Insurance domain based on Entity and Relation Extraction has significant practical implications.Based on the above background,this thesis has conducted extensive research on Entity and Relation Extraction,Knowledge Graph construction,and Question-Answering systems based on Knowledge Graph.The main research contents of this thesis include the following aspects:(1)In order to extract structured knowledge from unstructured text,this thesis first investigates methods for Named Entity Recognition and Relation Extraction.To address the lack of publicly available corpora in the field of Social Insurance,this study used Web crawling technology to retrieve a large amount of text related to Social Insurance from websites such as Baidu Baike.Then,a dataset for Named Entity Recognition and Relation Extraction in the field of Social Insurance was constructed.The article proposes a Named Entity Recognition model based on word-pair relation classification,which models two types of relations including the Next-Adjacent-Word(NAW)and the Tail-Head-Word＿*(THW＿*).to perform NER tasks.Subsequently,in relation extraction,after BERT encodes the input sentence,this thesis accesses Bi LSTM to obtain the semantic features of the sentence,and uses the semantic features of the sentence and the representation of two entities to perform feature fusion operations to improve the relation classification effect of the model.Through Named Entity Recognition and Relation Extraction,structured Social Insurance domain knowledge is obtained.Experiments have shown that the Named Entity Recognition model based on word pair relation classification and the relation extraction model combining sentence semantic features and entity features have achieved good results in Entity and Relation Extraction in the domain of Social Insurance.And the experimental results on public datasets show that compared with existing model methods,the proposed model in this article has certain improvements in three evaluation indicators: accuracy,recall,and F1 value.(2)This thesis uses the graph database Neo4 j to store the entity relation triplets extracted from the above research and visually display them in the form of a knowledge graph,and then we conducted an in-depth study on question-answering technology based on the Social Insurance domain knowledge graph.Knowledge graph-based question-answering technology mainly includes two parts: entity linking and relation prediction.In entity linking,this thesis calculates the similarity between entities through the Levenshtein Distance algorithm and proposes a multifeature fusion entity similarity calculation method based on this.In relation prediction,a relation prediction model based on a Siamese network is proposed.The model encodes the question and predefined relations using two identical network models and calculates the similarity between the question and relations to predict the relation in the question.After entity linking and relation prediction,the question generates a Cypher query statement and retrieves the answer in the knowledge graph.(3)Based on the above research,the design and implementation of a Social Insurance domain question-answering system is completed.This article uses system development technologies such as Python,Vue.js,and Django,and adopts a front-end and back-end separation approach to design and implement a question-answering system based on a knowledge graph in the Social Insurance domain.The system implements knowledge graph visualization,knowledge graph querying,entity relationship extraction,and automatic question answering functions.The experimental results in this study show that the Named Entity Recognition and Relation Extraction methods proposed in this thesis have certain advantages,and have practical application value in the field of Social Insurance;It is feasible to construct a knowledge map in the field of Social Insurance and apply it in a question-answering system.The research in this thesis has a positive role in promoting the construction and application of knowledge graph.