Research On Insurance Question Answering Algorithm Based On Knowledge Graph

Intelligent question answering is one of the core.In the context of the continuous development of the existing business in the insurance field and the continuous implementation of new business,the consulting services of potential customers and existing customers on the insurance business are showing an increasing trend,and higher requirements are placed on intelligent question answering methods.Traditional question answering methods are difficult to efficiently solve user problems and accurately understand semantic information.Deep learning methods do not have high-quality public Chinese insurance data.However,the natural language is diverse and complex,and the implementation of insurance intelligent question answering faces a series of problems.In order to realize the user insurance intelligent question and answer,the paper carry out in-depth research on the construction of insurance knowledge graph and question and answer based on knowledge graph,and through deep learning models and attention mechanisms to improve the accuracy of semantic expression,the research work is summarized as follows:(1)Proposing a construction method to construct knowledge graph in insurance field.It consists of three parts: knowledge extraction,knowledge fusion and knowledge storage.First,use Bi-LSTM-CRF and sequence labeling strategies to extract knowledge from the joint extraction model of entities and relationships;Secondly,a similarity calculation method of I-SPRS(Insurance-Surrounding Property and relation Similarity)is proposed to achieve knowledge fusion.Finally,a mixed method of graph database and relational database is used to realize knowledge storage.(2)A question and answer method for insurance knowledge graph is proposed,which is mainly composed of entity link and relationship prediction.Among them,the entity link is a design method of multi-feature fusion based on the characteristics of the insurance field.The relationship prediction is to build a model of a two-way gate control unit network,extracting the entities and relationships in the question.Use a searchable language to find and give the correct answer in the atlas.(3)Implementing the insurance knowledge graph question answering system,and realizing the knowledge graph visualization and build an insurance assistant.Improving the automatic question answering ability of insurance knowledge graph through deep learningmethod,Verifying the effectiveness of the algorithm model in this paper,and finally realizing an insurance service assistant that can satisfy the user's question and answer consultation.