Research And Implementation Of Structuring Processing Approach For Medical Semantic Understanding

With the boom of artificial intelligence,the application scenarios of artificial intelligence in the field of medical health are becoming more and more abundant,,and the technology of artificial intelligence affects the development of the medical industry.In some examination,doctors cannot leave the examination equipment with their hands,so it is urgent to introduce intelligent voice interactive medical products to assist work and improve work efficiency.Semantic understanding engine plays a core role in intelligent voice assistant.The meaning of medical semantic understanding is to help voice assistant understand doctors' intentions,extract the key information of doctors' speaking content,conduct structural processing on the obtained text information,and finally generate electronic medical records.Behind the flourishing development,the application and promotion of artificial intelligence in the medical field is also facing many problems and challenges.At present,there are many drawbacks in the structured processing of medical texts for Chinese natural language,such as insufficient flexibility,inability to realize the customization of various services,and easy loss of important medical record information.In view of the above problems,this paper mainly carries out the work from the following aspects.This paper is based on the "Medical Semantic Understanding Engine" project of the Wisdom Medical Core Division of IFLYTEK CO,LTD.,which studies the structured processing of voice transcribed text.In this paper.In this paper,a systematic structured processing scheme of "Rule + Named Entity Recognition + Knowledge Base + Classification" is proposed based on the in-depth analysis of speech transcribed text data and requirements.Firstly,aiming at the problem that the traditional information extraction technology is applied to this problem,this paper presents an information extraction processing method based on rule and named entity recognition fusion.The method performs NLP grammar analysis and extraction of named entity recognition.And keep the union of the extracted information.Secondly,the application of knowledge map in the structuring of traditional medical texts is only a simple splicing of semantics between entities,which has poor structuring effect.Therefore,this paper introduces the idea of knowledge map verification.The method is to verify the legality of the semantic information extracted in the structured system after the completion of the medical knowledge map,including the value type,the value range,and the semantic relationship,so as to improve the correctness rate of the text structure.Then,in order to prevent the loss of useful information in the text,this paper presents a method for classifying text based on CNN classification model,and improves the structure of CNN model.After experimental comparison and analysis,the text was classified into two categories using the combination of jieba participle and CNN.Finally,through the research of the first three main parts,a structured processing scheme for semantic understanding is designed and implemented.The improved system has improved the structure effect and classification effect through the real live voice transliteration text data verification.