Multi-Intent Text Classification Using Dual Channel Convolutional Neural Network
Zhiming Yang, Laiqi Wang, Yong Wang · 2019
During the conversations among human and machine, how to comprehend user's intention according to the context is a crucial issue, which is very difficult. The commonly used Convolutional Neural Network (CNN) does not work well in this situation: due to the short length and the local features of dialogue sentences, CNN cannot well capture the global features and the semantic information. In this paper, we propose an intent classification dual-channel Convolutional Neural Networks (ICDC NN): we first extract semantic features by using Word2vec and Embedding layer to train the word vector; then, two different channels are used for convolution, one for character level word vector, the other for word-level word vector; thirdly, the character level word vectors (fine-grained) are combined with word-level word vectors to mine more in-depth semantic information of natural language question; finally, with convolution kernels of different sizes, more in-depth abstract features inside the sentences are learned. We then apply the ICDCNN to English text but with a different network structure. Experimental results show that the algorithm achieves high accuracy on both English and Chinese datasets, which shows a significant improvement compared to other methods.