Classification and extraction of medical clinical trial screening standard texts based on Bi-LSTM and Attention mechanism

The medical recruitment of human subjects in clinical trials often requires manual comparison of the subject's medical records and clinical trial screening criteria, which is a time-consuming and labor-consuming method to determine candidates. In order to solve this problem, it has a great clinical value to study a method of classifying clinical trial data screening standards automatically. This paper attempts to propose a medical short text classification model of clinical medicine based on BiL-Att (Bi-LST M + Attention) model. It uses Word2vec prepossessing to get short text vector as the model input. The results showed that the classification effect of BiL-Att model reached t he highest Average F1 value of 80.26%.