Research On The Construction And Application Of Social Insurance Knowledge Graph

The social insurance system is closely related to people's livelihood,and social insurance as an important part of it is more inseparable from our ordinary people.However,with the development of the economy,a large amount of social insurance knowledge has become regionalized and fragmented,especially related to social insurance policies and regulations,and the development of the Internet has made people more willing to ask questions online.Therefore,it is urgent to build a knowledge base in the field of social insurance,and on top of it can create application value and solve the social insurance problems encountered by ordinary people,and the knowledge map can be used as a feasible method to solve the above problems.Based on the above background,with the support of natural language processing technology and neural network theory,this paper has carried out in-depth research on the construction of knowledge map and question and answer technology in social insurance field.The content of this paper is mainly divided into two modules: social insurance knowledge map construction and knowledge map based on knowledge map.In the knowledge map construction module of social insurance field,this paper proposes a knowledge map construction framework based on the characteristics of social insurance field,which mainly includes four parts: knowledge preprocessing,concept extraction,entity recognition and relationship extraction.In the preprocessing section,this article uses the web crawler to crawl the authoritative domain knowledge and denoise the text.Due to the many concepts in the social insurance field,in order to reduce the human participation,this paper first uses the combination of rules and statistics to extract the domain concept,and uses the constructed domain concept set for word segmentation and labeling tasks.In order to obtain the important components of the knowledge map: entities and relationships,this paper uses the classic Bi-LSTM-CRF entity recognition model to identify the entities in the social insurance field,and improve the characteristics of the basic domain,adding the word segmentation layer.And use combination embedding to improve recognition.Next,this paper uses the method of remote supervision to design the social insurance domain relationship extraction framework,using multi-segment convolutional neural networks(PCNNs)to model,and introduce attention mechanism and entity description features.The domain knowledge map is then constructed from the extracted entities and relationships.Based on the constructed knowledge map,this paper studies the question and answer technology in the field of social insurance,which is mainly composed of two tasks: entity link and relationship prediction.In the entity link part,this paper proposes a method for calculating the similarity between entities with multiple features.Then in the relationship prediction task,this paper constructs a relational prediction model based on cascading Bi-GRUs and introduces the attention mechanism.Through the above two steps,the entities and relationships in the question can be extracted,and finally form a formal query statement,and the result is returned after querying in the knowledge map.Finally,this paper experiments and analyzes the methods and models proposed or adopted above to verify the validity and accuracy of the proposed algorithm and model,and illustrates effectiveness and application value of the method of constructing the knowledge map in social insurance field.