A Review of Deep Learning Models for Twitter Sentiment Analysis: Challenges and Opportunities

Abstract

Microblogging site Twitter is one of the most influential online social media websites, that offers a platform for the masses to communicate, express their opinions, and share information on a wide range of subjects and products, resulting in the creation of a large amount of unstructured data. This has attracted significant attention from researchers, who seek to understand and analyze the sentiments contained within this massive user-generated text. The task of sentiment analysis entails extracting and identifying user opinions from the text, and various lexicon and machine learning-based methods have been developed over the years to accomplish this. However, deep learning-based approaches have recently become dominant due to their superior performance. The current study briefs on standard preprocessing techniques and various word embeddings for data preparation. It then delves into a taxonomy to provide a comprehensive summary of deep learning-based approaches. Additionally, the work compiles popular benchmark datasets and highlights evaluation metrics employed for performance measures as well as the resources available in the public domain to aid sentiment analysis tasks. Furthermore, the survey discusses domain-specific practical applications of sentiment analysis tasks. Finally, the study concludes with various research challenges and outlines future outlooks for further investigation.