| As human society enters the Internet era,all aspects of people’s lives are closely related to the Internet.By June 2021,the number of Internet users in China has reached 1.011 billion.In today’s information age with a huge amount of data,people have higher requirements for the speed and efficiency of finding information.To improve the efficiency of information retrieval,an intelligent question answering system came into being.Most of the existing medical Q＆A system uses doctors’ online consultations.There are some problems,such as patients needing to wait for a certain time,delayed communication with doctors,expensive consultations,limited doctor resources,and so on.At present,when extracting medical entities,the intelligent question answering system for the medical field has the process of missing extraction and lack of medical entity audit at the back end of the system,which leads to the mixing of traditional Chinese medicine data and non-medical data in the database,greatly reduces the question answering efficiency and does not achieve good results in the field of intelligent medical question answering.There are some problems in offline treatment,such as the difficulty of obtaining expert number one in well-known hospitals,the queuing time accounting for more than 75% of the treatment time,and the ambiguity of the corresponding department of patients.At the beginning of 2020,due to the outbreak of novel coronavirus pneumonia,hospitals made stricter regulations on the medical treatment of patients,which made it more difficult for patients to seek medical treatment.In this case,people more hope to query their diseases and seek treatment through a professional,efficient and intelligent medical Q＆A platform.Therefore,it is urgent to build an intelligent question answering system for the medical field.With the development of deep learning technology,we combine the deep learning method with the medical question and answer system,use the deep learning method to extract and classify the medical data quickly and accurately,and form a pure database for the medical field,to improve the question and answer efficiency and solve the problems of low question and answer efficiency and long waiting time caused by manual inquiry.To alleviate the pressure of medical treatment and meet the needs of users for online medical treatment,an intelligent medical question answering system based on a cyclic neural network is developed in this paper.Because of the poor ability to parse natural language in the existing medical Q＆A system and the lack of review of medical entities at the back end of the system.In the back-end data processing part of the question and answer system,this paper uses the medical entity recognition model based on BERT-Bi LSTM-CRF to recognize the medical entity of unstructured medical data.This paper adds a medical text classification module at the back end of the system to audit medical entities.In this module,the medical text data and non-medical text data are classified by using the medical text classification model based on the BERT-RNN proposed in this paper.To improve the accuracy and speed of model classification,the mean proposed in this paper is used mean＿relu activation function replaces the original activation function of RNN and adjusts the network structure of RNN,to realize the efficient classification of medical data.Due to the particularity of medical words,the same word may represent different disease names in different positions in the sentence.Therefore,the word vector transformation part of the above two models is realized by BERT.Through the BERT word vector transformation method,the word vector,text vector,and position vector information in the sentence are fused,so that the word vector can learn not only all the semantic information of the word itself but also the relationship between other words in the sentence and the word.In this way,the medical question and answer system can obtain more accurate medical data expression from the root,which is conducive to the realization of the system’s function.Finally,by comparing the classification accuracy and classification speed between the BERT-RNN medical text classification model proposed in this paper and the existing BERT-LSTM text classification model under the same data set and the same number of iterations,it can be obtained that the classification accuracy of the BERT-RNN medical text classification model proposed in this paper can reach 91.2%,which is improved by5.9% compared with the BERT-LSTM text classification model.In terms of classification speed,in the same case of 10000 iterations,the classification speed of BERT-RNN medical text classification model proposed in this paper is 30m46 s,which is 1m45 s higher than that of the BERT-LSTM text classification model.Firstly,this paper analyzes the requirements of the intelligent question answering system in the medical field.According to the system requirements analysis,it is determined that the system designed in this paper adopts B/S architecture.On this basis,the overall framework,overall functional structure,and database involved in the system are designed.Finally,the system development framework and the software and hardware environment required to run the system are configured accordingly.After setting up the environment required by the system,realize the functions of the system.Through the test,it is verified that the function and performance of the system have achieved the expected effect,the system can realize various functions in the design,and each module has reached the ideal state.The intelligent medical question answering system based on the cyclic neural network can answer the questions raised by users in real-time,improve the utilization of medical resources and the accuracy of question answering,and meet the needs of users for a medical question answering system.Therefore,the intelligent medical question answering system based on the cyclic neural network developed in this paper has a good application prospect. |
PDF Full Text Request