A Comprehensive Categorization and Comparative Analysis of Methodologies in Aspect-Based Sentiment Analysis

The increasing number of user-generated reviews on the web has forced businesses to want to know more about the sentiment of customers, which has led to the need for improved techniques in the field of sentiment analysis. Aspect-based sentiment Analysis (ABSA) addresses this need by extracting detailed sentiment information linked to specific aspects or categories in text. This survey presents an overview of the current achievements in ABSA, which suggests a new taxonomy that is used to categorize ABSA methods into three main groups: machine learning and deep learning-based models, attention-based models, and transformer-based models. The taxonomy is focused on how the four key components of ABSA (aspect, sentiment, category, and opinion) have interacted. Key models are explained with technical details and in simple and intuitive descriptions, along with a comparison of their performance. It also outlines the current challenges and emerging trends to offer practical guidelines that can guide the direction of future research and applications in ABSA.