Sentiment analysis based on deep learning approaches for text classification

Recently, text-based sentiment analysis has been a hot area of research. The automatic identification of opinion words and their sentiment polarity is among the most notable efforts. Despite the success of sentiment analysis for text, the methodological issues are understudied. In this direction, there are several significant setbacks regarding datasets, methodologies, baselines, statistical analyses, and comparisons between information from various sources. Using a well-known dataset, this paper experiments with three advanced transformer-based deep learning techniques: ALBERT, RoBERTa, and VADER. A performance evaluation matrix based on accuracy is used to compare the performance of these models on the dataset. Based on the results, RoBERTa and ALBERT performed better than VADER, respectively, with 86%, 87%, and 83% accuracy. To improve sentiment classification performance, future research could improve the models’ architecture.