Comparative Analysis of Deep Learning Approaches for Aspect-Based Sentiment Classification

Social networking-based platforms (e.g., Twitter, Facebook, Amazon) where users feel free to express their opinion/sentiment about any product, service or event, in the form of reviews, tweets, blogs, etc., in the last decade with the enhanced use of social media and social networking, result in huge amount of data in unstructured and unorganized form, and the need of sentiment analysis of user’s opinionated text arises for improvement in the product or service policymaking decisions. Therefore, sentiments in user’s text play crucial role, and classifying the sentiment (positive, negative or neutral) at review/document level or at sentence level is not sufficient for achieving the real solution for improvement in product, service or an event; therefore, there is a need of deeper level and fine-grained analysis at aspect level known as aspect-based sentiment classification (ABSC) which attract researchers. This paper emphasizes on various existing approaches for ABSC specifically deep learning methods and their comparative analysis.