Research on Case Preprocessing Based on Bert -CNN-LSTM Model

Abstract

In this paper, we apply the deep learning algorithm to preprocess the criminal case information. By extracting the characteristics of the brief case field, the missing content of other required field is plugged through model training. The experimental results show that the precision rate of CNN-LSTM model is 3% higher than that of LSTM-CNN model in text classification. After the Bert model is integrated, the precision rate, recall rate, and F value are all improved by 10%. To the best of our knowledge, this is the first time to use Bert model in preprocessing criminal case information.