Electronic medical record text classification method

Abstract

The invention discloses an electronic medical record text classification method. The method comprises the following steps: performing preprocessing operation on an original electronic medical record text data set to form an original corpus, converting the original corpus into a word list T1, training the word list T1 by utilizing a word vector tool, expressing word training as low-dimensional dense word vectors, and forming a word list T2; and then, correspondingly converting each piece of data of the text data set into a word vector sequence as input in a word number form, using a two-channel structure of a CNN-Attention neural network and a two-channel structure of a BiLSTM-Attention neural network for training a text feature vector, then splicing output of the two-channel structure to serve as overall output of the neural network to obtain the overall output of the neural network, and finally, calculating the probability of the tag category to which the text belongs by using a softmax classifier. According to the method, local and global text features of the electronic medical record text data set can be planned as a whole, good stability and robustness are achieved, and the effect of an electronic medical record text classification model is effectively improved.

Description

technical field [0001] The invention relates to a classification method, in particular to an electronic medical record text classification method, which belongs to the technical field of application of natural language processing to medical electronic cases. Background technique [0002] Text classification refers to establishing a relationship model between text and categories. As one of the basic tasks of natural language processing, it is of great significance in sentiment analysis, social platform public opinion monitoring, and spam identification. The main algorithm models of text classification can be basically divided into three categories: the first category is based on rules, the second category is based on statistics and machine learning, and the third category is based on deep learning methods. [0003] The first type of rule-based method relies on the help of professionals to formulate a large number of judgment rules for predefined categories, and the degree of ...

AI Technical Summary

Problems solved by technology

At present, the text classification of electronic medical records using the neural network combination model is limited by the high-dimensional and sparse text features of electronic medical records, dense text terms, missing sentence components, etc., which will cause slow model convergence and poor classification results.

Images