Weibo Comments Sentiment Analysis Based on Deep Learning Model

Xixiang Hu,Yu Zhang,Hongli Zhang
DOI: https://doi.org/10.2991/caai-17.2017.119
2017-01-01
Abstract:In this paper, the sentiment analysis based on the deep learning model was studied. By comparing the effects of the shallow learning model SVM and the deep learning model LSTM on the classification of Weibo comments, we found that the classification result of the LSTM model is better than that of the SVM model under the same group of training set and test set. One of the main reasons lies in that the LSTM model exploits the information of word order during training. Firstly, we need to crawl the original Weibo data from Website and pre-process these data. Secondly, with the help of the neural network language model inside word2vec, the word embedding as the input of the SVM model and the LSTM model are trained. Thirdly, inputting the training set and test set constructed by the word embedding into the SVM model and the LSTM model respectively, then we got the experimental results. Finally, by comparing the experimental results of the SVM model and the LSTM model, we found a way to improve the accuracy of sentiment classification by introducing the information of word order in the model.
What problem does this paper attempt to address?