Evolving Methods in Social Media Sentiment Analysis: Innovations and Challenges

Abstract

Several major studies conducted during the period 2010-2023 have been compiled in this article to illustrate recent progress made toward improving sentiment analysis methods [1] applied to social media websites during that period. In light of the vast amount of user-generated content that is uploaded daily to social media platforms[2], [3] such as Facebook or Instagram, it is essential to evaluate public sentiment on these channels when marketing or monitoring public opinion processes. As a result, traditional machine learning models[4] cannot comprehend the unstructured and evolving language found on social media platforms. For the purpose of improving accuracy, researchers have developed ensemble learning techniques[5] as well as hybrid approaches [6]. CNNs, RNNs, and Transformer models [7] such as BERT have revolutionized sentiment analysis by revealing intricate details. An analysis of real-time sentiment in social media videos can be quickly performed using big data analytics. Multimodal methods [8], [9] incorporating visual, audio, and textual data provide a more comprehensive understanding of sentiments in social media videos. Model adaptability across datasets is improved through transfer learning. Unfortunately, differences in language and ethical issues remain, emphasizing the need for ongoing research to develop adaptable, expandable, ethical sentiment analysis models. The purpose of this review is to discuss the ground-breaking implications of advanced techniques for analyzing social media sentiment, along with future research directions.