An analytical assessment of sentiment analysis trends and methods through systematic review and topic modeling

Sentiment Analysis (SA) applies Artificial Intelligence (AI) and Machine Learning (ML) techniques to identify and interpret opinions, emotions, and sentiment polarity in text data. This study presents a systematic literature review of SA research published between 2012 and 2024 using a dataset of 14,482 journal articles indexed in Scopus. The primary objective of this research is to provide a data-driven overview of the evolution of SA research, offering insights into methodological evolutions, emerging application areas, and the growing influence of AI in the field. The research findings reveal a multidisciplinary growth of SA research, with increasing contributions from Health Sciences and Physical Sciences areas, alongside traditional domains such as Computer Science and Engineering. Author keyword trends highlight a methodological shift from lexicon and ML-based approaches towards AI and Deep Learning (DL) techniques, with the rising prominence of models such as CNN, LSTM, and BERT. We also employ topic modeling and identify that SA's significant methodological and application themes include business, public health, education, and social media.