Research On Audit Interactive Encyclopedia And Knowledge Question Answering System

The audit interactive encyclopedia and Q&A system aims to build a Chinese sharing platform for auditing knowledge.Through this platform,the audited personnel can fully understand the basic situation of the audit in the field,so as to find their own problems in the daily work,correct it in time,and avoid making a bigger mistake,at the same time the auditor can use the platform to increase their knowledge reserves and more efficient in their work.This paper takes the field of medical insurance audit as the research object to study how to realize the question and answer of the knowledge in the field of medical insurance auditing.It is mainly divided into two parts: the construction of knowledge graph and the analysis of user query questions in the field of medical insurance auditing.In the construction of the knowledge graph,using the crawler technology to acquire relevant knowledge from the Internet,construct the domain dictionary,and serve as the basis for the construction of domain knowledge graph.Then,we focus on extracting domain entities from unstructured text and an entity recognition algorithm for fusion domain is designed to identify domain entities.This paper learns the idea of deep learning and the innovative way of expressing features in the design domain.It uses the distributed word vectors to express the semantic information of words,and uses the similarity between domain candidate words and the word vectors of the domain dictionary to characterize the domain.feature.After that,some lexical rules are summarized in the field dictionary to further improve the effect of domain entity recognition.Finally,use the graph database Neo4 j to store the acquired knowledge.For question parsing this question,this paper draws on the idea of the maximum forward matching segmentation algorithm and uses the entity set in the knowledge graph as a dictionary to extract entities in the question sentence.Then,a set of candidate attributes of the entity is obtained according to the entity query knowledge graph.To get(entity,attribute,?)triples,build a Cypher query for the user to return the answer.The basic LSTM model is improved to deal with the implicit attributes in user query questions and the corresponding attributes in the knowledge base.Specifically,in the implicit attribute mapping of question sentences,this paper attempts to use the bi-directional LSTM,and designs a word-by-word attention mechanism to improve the effect of the implicit attribute mapping of the question sentence,giving the user a better Q & A experience.This paper designed a comparative experiment to verify the effectiveness of the proposed algorithm and the improved model designed in this paper.This shows it is feasible to construct question answering system in the field of medical insurance auditing,which provides a good idea for the construction of the audit interactive encyclopedia and knowledge question answering system.