Design Of Intelligent Traditional Chinese Medicine Question Answering System Based On Deep Learning

The re-emergence of neural networks has brought tremendous impact to the societ y.The most in-depth combination of application and image processing and natural language processing,the application of deep learning chat robots is one of the hot spots.In recent years,the combination of chat robots and professional fields has become the rese arch direction of the current ver y hot artificial intelligence.At present,researchers all over the world are using deep learning technology to develop s ystems when researching and developing intelligent chat robots for professional fields.At present,man y scholars try to design a system b y combining retrieval and generation.The research in this paper is based on this.Ideas are designed.This paper anal yz es and studies the development of chat robots and their applications in the medical field.Combining with the current development of Chinese medicine,this paper proposes to design a learnable intelligent question answering system that can be used for TCM consultation.The ke y work done in this paper is as follows: 1.Through the study of word embeddin g technology,learn the application of word vector trained in Google’s Word2 Vec training tool in depth model,the performance of experimental word vector and unique heat coding in processing text;Accuratel y search the medical inquir y data set required for the development of the positioning s ystem,collect and mark the data sets needed for s ystem design b y stud ying the web crawler technology,stud y the basics of natural language processing,and effectivel y process the s ystem data set,including establishing a term dictionar y and saving The effective data after sorting;3,through the experimental SVM and convolutional neural network performance on the text classification dataset,research and improve the model,using the convolutional neural network and SVM to combine the model to classif y the user’s questions,The classification accurac y rate is improved.4.The neural network based on the bidirectional long-term and short-term memor y(BLS TM)network is applied to the user statement,and the semantic feature information of the sentence context is fully extracted,which improves the correct answer of the sentence;5.Studying the sequence-to-sequence model theory,modif ying the encoding and decoding part of the optimized Seq2 Seq model for the corpus data used i n this paper.The coding stage of the model uses the bidirectional long-term and short-term memor y network to extract more comprehensive sentence information coding,while decoding The stage uses a bidirectional long short-term memor y network,and uses the attention mechanism to improve the effectiveness of the sequence generation statement by optimizing the number of la yers of the neural network.6.Designing the user interaction interface,completing the user registration and login-free design,and design ing the switchover.Mode function,the user passes the function Selectivel y retrieving matching s ystem back to view the model or system generated reply intelligent generated.