Research On The Construction Method Of Knowledge Graph For Medical Insurance Intelligent Audit

Medical insurance is an important part of social insurance,which is of great significance to the people.The implementation of the medical insurance system is inseparable from the support of the medical security fund.However,restricted by the imperfect monitoring system and the low degree of supervision informatization,the insurance fund fraud often occurs,and the fund supervision situation is more severe.The state attaches great importance to it.It proposes to establish an intelligent monitoring system in an all-round way,enhance the ability of intelligent supervision of medical insurance,and actively explore new mode of "Internet plus medical health",so as to transform the fund supervision into specialization,informatization and intellectualization.The realization of medical insurance intelligent audit system needs to establish the knowledge base related to the medical field.Because the knowledge graph can efficiently query the related data,it is suitable for the situation of large amount of data and more connections between data.Therefore,this paper uses the knowledge graph to establish the knowledge base.The knowledge graph is composed of three tuples.The three tuple extraction task can be divided into the task of identifying entities and the task of extracting relationships.In the first task,based on the disadvantage that the LSTM model may forget the historical information associated with the current information,this paper proposes an improved model RFLSTM,which first fuses with the new information before the old information is forgotten,and then passes through the forgetting gate.Combined with the attention mechanism and the pre training language model BERT,this paper proposes the BERT + birflstm + CRF model.In the relationship extraction task,in order to avoid the problems of error accumulation and relationship overlap in the pipeline model,this paper adopts the entity relationship joint extraction scheme.In order to identify not only triples but also entity types,this paper proposes a multi-layer pointer annotation strategy,and further proposes the entity relationship joint extraction model MLPER based on this annotation strategy,In addition,this paper adds the shared coding layer,head entity vector,head entity type vector,comparative tail entity vector,comparative tail entity type vector and the comparative position vector between each word and head entity to the vector splicing layer of the model,so that the model can obtain more information.The experimentals show that the performance of the proposed RFLSTM and MLPER is improved.Finally,taking the national prescription set and medical insurance drug catalogue as data sources,the medical insurance audit knowledge graph is constructed by using MLPER model,and the medical insurance data is audited based on the constructed knowledge graph.The results show that the construction method of medical insurance intelligent audit knowledge graph proposed in this paper is effective.