A Systematic Literature Review on Multimodal Aspect-Based Sentiment Analysis

Multimodal Aspect-Based Sentiment Analysis (MABSA) is an emerging field that enhances traditional sentiment analysis by incorporating different data types like text, image, audio, and video. Given the range of data available today, MABSA aims to offer a comprehensive understanding of sentiment by integrating multiple modalities. This systematic literature review compiles and critically analyses primary studies to address key research questions in MABSA, offering a detailed examination of current methodologies, datasets, evaluation metrics and approaches. We extract and analyze information on techniques, datasets, modality representations, and performance metrics to identify strengths, weaknesses, and influencing factors. Four research questions guide this review, focusing on the evolution, effectiveness, and future directions of MABSA research. The findings reveal that most studies utilize transformer-based approaches, and tackle various challenges such as addressing the issue of semantic gaps.