Research On Clinical Medical Text Classification Based On Hybrid Deep Learning Model

With the advancement of the "Double First-rate" construction strategy,large public hospitals increasingly attach importance to discipline construction.In particular,the geographical advantages of hospitals are gradually reduced,and the pressure on competition is increasing,so it has become an inherent need to promote the development of public hospitals that strengthens the construction of disciplines to improve the level of medical treatment,scientific research,and teaching,and thus improves the core competence of hospitals.Discipline evaluation guide the direction of discipline construction and evaluate the results of discipline construction,and it’s also an important starting point to improve the core competence of hospitals.In discipline evaluation,text classification of text data such as SCI papers is a basis work,the accuracy and efficiency of text classification have a great impact on the smooth progress of disciplinary evaluation.At present,there are few published studies on text classification in discipline evaluation.In the practice of text classification for discipline evaluation,the STEM research launched by the Chinese Academy of Medical Sciences has used the vocabulary matching and other text classification methods to carry out the subject classification of text data in clinical medicine.However,the vocabulary matching method has problems such as difficulty in constructing the subject vocabulary and high time cost.The text classification algorithms based on Machine learning have problems that don’t consider the semantic relationship of words,loss of classification information during feature engineering,and insufficient computing power of the model.These problems lead to poor overall classification performance of text classification and misclassification of similar disciplines in STEM research.Considering that deep learning methods have better performance on news corpus and social website review corpus than machine learning-based text classification methods,this article introduces deep learning related models in the text classification research of discipline evaluation and builds a clinical medical text classification model based on the hybrid deep learning model.The main work of this article includes:(1)Design and implement a clinical medical text automatic classification model.This research introduce the ideas and methods of deep learning in the text classification of discipline evaluation.In text representation stage,word vector is used to represent the text,and in the classifier construction stage,two deep neural network models are used to construct a hybrid model.Using SCI paper data in STEM research to carry out empirical analysis,it is confirmed that in the text classification practice of discipline evaluation,the classification performance of the model is nearly 7 percentage points higher than that of the traditional SVM model.And as for the three pairs of disciplines with overlapping research contents and easy to be misclassified,the performance of this model is nearly 8 percentage point higher than that of the traditional SVM model.(2)Construct a text classification prototype system for medical discipline evaluation.This research builds a text classification prototype system based on the model constructed in this research.The prototype system can provide automatic classification and manual review services for SCI papers,as a preliminary exploration of the application scenarios of this study in the text classification practice of discipline evaluation.