Questions and Answers on Legal Texts Based on BERT-BiGRU
Nana Zhang, Yinan Xing · Journal of Physics Conference Series · 2021
Abstract Using text question and answer technology to improve the efficiency of judicial personnel in the process of case handling can greatly reduce labour costs. This paper proposes a hybrid neural network model that combines pre-training model BERT and Bi-GRU. The model first uses the pre-training model to learn powerful semantic capabilities, then combines Bi-GRU to learn the semantic information between the text and the question, finally gets the answer corresponding to the legal text. The experimental results on the CJRC data set show that compared with the basic baseline model, the algorithm in this paper can effectively improve the accuracy and F1 value.