Text Sentiment Classification Based On Deep Learning

The text sentiment classification has occupied a very important position in the study of sentiment analysis.Now with the explosive growth of information,many researchers have applied the theory of deep learning to text sentiment classification,especially the neural network model has achieved good results in sentence and document sentiment classification.Convolution neural network can capture the local relationship of spatial structure,but lack the ability of learning the relationship between continuous sentences.Recurrent neural network can deal with the continuity information,but in parallel to extract the text features are powerless.Aiming at these problems,this study constructs a new deep learning model C-LSTM(Convolutional-Long Short-Term Memory)for text sentiment classification,which combines the advantages of convolution neural networks and recurrent neural networks.The specific work is as follows:1.Study the current mainstream deep learning model for text sentiment classification,such as convolution neural network and recurrent neural networks.2.Construct the Chinese text sentiment classification model based on C-LSTM.The model uses the convolution neural network to extract the high level word features,and the result is input to LSTM to get the representation of the sentence,which not only obtains the local features and obtains the global semantic features.3.Collect 25,000 Chinese text data with emotional markings in six fields through the way of web crawler for the research.4.Train the model to achieve the best results,and compare the experimental results of the three models,the C-LSTM model do improve the classification accuracy.