A Taxonomy for Sentiment Analysis on Social Networks: AI Approaches and Synthesis

As social network grew rapidly, people now communicate more through needs, opinions, and feelings in posts and comments, generating massive dynamic data daily. Appropriate processing and analysis of such data are now vital in understanding user expectations and making sound decisions. Social media Sentiment Analysis is, nonetheless, hampered by various challenges like linguistic ambiguity, sarcasm detection, and scalability. In this paper, we present a comprehensive survey of SA techniques for social networks. We propose a structured taxonomy classifying existing methods based on major challenges, employed techniques, employed datasets, and metrics for evaluation. By the comparison of merits and demerits of various Machine Learning (ML) and Deep Learning (DL) models, this paper highlights the trade-offs between model performance, computational expenses, and interpretability. Our results present some insights about the state of sentiment analysis on social networks and indicate some areas for future work.