Design And Implementation Of Intelligent Question Answering System For Medical Field

With the continuous improvement of national health awareness,people pay more attention to medical and health knowledge,and put forward higher requirements for the way to obtain medical and health knowledge.At present,the medical knowledge acquisition methods provided on the Internet are still in the stage of relying on keyword search and artificial reply,and the lack of convenience,timeliness,pertinence and intelligence of information acquisition affects the popularization of medical knowledge.At the same time,the emergence of a new generation of natural language processing technology represented by deep learning technology makes the intelligent question answering system more mature.Therefore,aiming at people’s demand for convenient access to medical knowledge and combining with the increasingly mature natural language processing technology,this paper studies and implements an intelligent question-and-answer system for the medical field.In response to users’ questions,this paper uses the techniques of entity recognition,text classification and semantic similarity calculation to process and analyze user questions,obtains the knowledge corresponding to questions in the medical knowledge base,and generates the answers by stitching or obtains the answers matching to user questions in the question and answer database.Firstly,this paper constructs the medical knowledge base with disease as the core by using the web crawler technology,and USES the Neo4 j graph database storage scheme to realize the structured storage of medical knowledge.Secondly,the design and implementation based on entity recognition and semantic parsing function of text classification,in terms of entity recognition,through the integrated use of medical dictionary matching entities and entity recognition model based on BiLSTM + CRF,and use the rules and the method of edit distance calculation to determine the results in fusion processing,effectively improve the effect on the entity recognition in medical entity vocabulary;In terms of text classification,the BiLSTM text classification model based on fusion features is designed and implemented,and the entity recognition result is extracted as onehot vector and integrated into the model,which improves the effect of the model in medical questions classification.In addition,a sentence vector generation method based on tf-idf weighting and word2 vec is implemented,and the semantic similarity is calculated by using cosine similarity.The word2 vec word vector is weighted by the tf-idf algorithm to improve the proportion of core words in the sentence vector,and the good effect of similarity calculation for question matching is obtained.Finally,the answer generation method based on template matching and knowledge base query and the answer generation method based on semantic similarity calculation result query set retrieval are used to obtain the question answer.The test results show that the intelligent question-and-answer system for medical field implemented in this paper can correctly understand the content of users’ questions and give professional answers in line with expectations.The system performs well,and the throughput rate and response time are all within a reasonable range.