An integrated neural model for sentence classification
Yanbu Guo, Weihua Li, Chen Jin, Yunhao Duan, Shuang Wu · 2018
Text classification is one of the fundamental problems in natural language processing. The difficulty of text classification still lies in the ambiguity and rhetorical of natural language. In order to classify sentences more effectively, we propose a hybrid L-MCNN model to represent the semantics of sentences, and this model consists of two parts, bidirectional long short-term memory (LSTM) and multichannel convolutional neural networks (CNNs). Different from the existing methods, our model utilizes the bidirectional LSTM to acquire the max contextual information of every position in sentences, and then the abstract semantics would be fed into the multichannel convolution layer which uses several filters to extract active local n-gram features. By combining the advantages of both architectures, the L-MCNN model can capture both long-distance dependencies and local information within sentences to improve the classification performance. Several experiments are conducted, and the results indicate that our system outperforms the baseline algorithms on sentiment analysis and question classification tasks.