Chinese Intention Recognition Algorithm Based on BERT-GRU-Capsule Network

Abstract

With the continuous development of natural language processing technology and deep learning technology, related technologies in intelligent question answering field have made rapid progress in recent years. In the application of intelligent question answering system, the research of intention recognition algorithm is very important. In fact, intention recognition corresponds to the task of multi-classification of short texts in the field of natural language processing, and the algorithm of intention recognition is directly related to the question answering effect. In this paper, a BERT-GRU-Capsule network is proposed and compared with some classical intent-recognition networks on the user intent-domain classification dataset of SMP2017-ECDT.The experimental results show that the precision, recall and f1 values of the proposed network on the test set reach 0.931, 0.925 and 0.926 respectively, and the experimental results prove that the proposed Chinese intention recognition algorithm is superior to the classical intention recognition algorithm.