Design And Implementation Of Medical Information Retrieval System Based On Text Mining

With the improvement of material,culture and living standards,people pay more and more attention to their own health.However,traditional offline medical services have many inconveniences such as time and space.With the rapid development of informatization and intelligence,more and more people have the need to read medical popular science information and obtain relevant knowledge online.However,the current general search engine is difficult to guarantee the reliability and accuracy of search results when searching for information in vertical fields,especially in the medical field.The information retrieval platforms for the medical field are mostly for the integration and display of raw data.In order to provide reliable,accurate and fast medical information retrieval service,the medical information retrieval system is designed and implemented.The medical information retrieval system consists of three modules: data processing module,question answering system module and information retrieval module.The data processing module is used to crawl and integrate the basic data;the question answering system module is to process the text data through a series of algorithm models to provide question-and-answer matching services;the information retrieval module is to build a platform for users with data retrieval services and other general services.The core data source of the system is crawling.In order to ensure the accuracy of medical science popularization information,this article uses a manual pre-selection method to collect and extract portal data,and stores structured and semi-structured target data separately through ETL processing into My SQL and HBase.In order to solve the errors and gaps in the data of major portals,the author proposes a data integration strategy based on the priority of the data source.After the data is integrated,it is imported into Elastic Search to build a distributed cluster,and batch updates of Elastic Search indexes are implemented.The web platform uses a microservices architecture and uses the currently popular framework Spring Cloud to provide users with general retrieval,article retrieval,question-and-answer retrieval and other retrieval services,as well as general services such as collections,reviews,and likes.The author proposes a FAQ questionnaire model based on sentence-pair similarity calculation task,combined with named entity recognition task and intent recognition task to preliminarily screen the candidate set,and achieves a balance between accuracyand timeliness of the question answering system.Among them,the medical entity recognition uses the BERT-Bi LSTM-CRF model,the intent recognition uses the Text CNN model,and the sentence pair similarity calculation uses the fine-tuning BERT model.The medical information retrieval system brings great convenience to users,and users can focus more on information retrieval and knowledge acquisition in the medical field.The platform will provide users with intelligent and professional information retrieval services according to user needs.At present,the system has passed the functional test and non-functional test,waiting to be deployed on-line.