Hierarchical convolutional recurrent neural network for Chinese text classification

Text classification is the most fundamental and essential task in natural language processing. The last decade has seen a surge of research in this area due to the unprecedented success of deep learning. Previous models assume that all classes are equally difficult to distinguish and treat all of them exclusively. But in fact, the property of general-to-specific category ordering often exists between classes. In this paper, we exploit the prior knowledge of class hierarchical structure to enforce the network to learn human-understandable concepts in different blocks and propose a new model named H-CRNN, which combines TextCNN and Bi-LSTM to construct a hierarchical structure. We test our proposed model on the THUCNews dataset, and experiments show that our proposed H-CRNN model achieves the best results than other methods.