Deep Learning Framework Study for Twitter Sentiment Analysis

Nowadays social media shows surging significance in our lives. Sentiment analysis emerged to automatically predict the sentiment expressed by the comments on platforms such as twitter and Weibo. In this paper, a number of classical machine-learning-based methods are introduced, discussed, and evaluated at first. Then, a deep-learning-based model for sentiment analysis is proposed. The main contribution consists of avoiding underfitting and overfitting, dropout, different layers exploration, and LSTM. In addition, a publicly available benchmark dataset is utilized for evaluation. Finally, variety of evaluation methods are used for the proposed model. The results show promising results against other classical machine learning models.