Short Text Sentiment Classification Based on CNN-LSTM Model

A CNN-LSTM model-based short text sentiment classification method was proposed to effectively obtain the implicit semantic information of short text reviews. The convolutional neural network (CNN) model was used to automatically learn the semantic feature by setting different sizes of convolution windows. The long short-term memory (LSTM) neural network model was used to predict the sentimental label of the short text. The performance of the model was evaluated on three different short text review data sets. The F₁ value of the positive and negative data in NLPCC is 0.768 3 and 0.772 4, respectively (better than the best NLPCC evaluation result). Compared with the traditional machine learning classification model, t-test results show that the performance is improved significantly.