Design And Implementation Of Medical Question Answering System Based On NoSQL

Today,people pay more and more attention to medical and health problems.With the advent of the Internet era,more and more users are eager to obtain medical knowledge from the Internet.However,although there are a lot of medical information spread on the Internet,it is difficult for ordinary Internet users to obtain and use the information effectively through search engines.To solve this problem,this thesis designs and implements a medical Q&A system based on NoSQL(Not Only SQL),so that users can quickly obtain their desired medical knowledge through questioning,and thus realizes the effective use of medical data on the Internet.At present,there are few Q&A systems in the medical and health field,so the medical Q&A system designed in this thesis has a wide application prospect.The medical knowledge graph based on NoSQL can also be applied to more application scenarios such as medical chat robots in the future,and further promote the process of intelligent medical system.This thesis first proposes a data storage scheme based on NoSQL to solve the problem of various entities and relationships of medical data.It uses the graph database Neo4 j to build a medical knowledge graph,which reduces the difficulty of medical data storage.Next,aiming at the problem of how to recognize named entities from users’ questions,this thesis proposes a named entity recognition model based on Transformer encoder,which improves the precision of named entity recognition task on two Chinese datasets by adding dictionary information on the char embedding layer and adding relative position information on the Transformer encoder layer.Then,aiming at the problem of how to find the most similar relationship in the medical knowledge graph with the user’s semantics,this thesis proposes a semantic similarity calculation model based on Bi GRU(Bidirectional Gated Recurrent Unit),which uses Bi GRU instead of Bi LSTM(Bi-directional Long Short-Term Memory)to improve the training speed of the semantic similarity calculation model,and adds the interactive attention mechanism to extract the local similarity features between sentences.Experimental results show that the proposed semantic similarity calculation model has higher precision and faster convergence speed.Finally,on the basis of the above work,the medical Q&A system based on NoSQL is implemented and tested,which meets the needs of users to obtain medical knowledge through simple questions,and realizes the visualization of medical data.