Advancements in Aspect-based Sentiment Analysis: Leveraging Deep Learning Alogrithms for Accurate Sentiment Classification
Manish Kumar, Vikas Mangotra · 2025
Sentiment analysis (SA) functions as an NLP sub area which provides the capacity to analyze textual subjectivity that occurs extensively on digital platforms including social media and e-commerce and customer surveys. These platforms grow, and so too has the need to accurately gauge public sentiment become more and more important to businesses and researchers as well as policymakers. The article covers sentiment analysis development from its initial lexical examination phase into its present-day advanced ML and DL techniques. The problem we study is aspect-based sentiment analysis (ABSA), which is to classify sentiment regarding particular aspects of entities, which provided finer grained information than sentiment polarity classification alone. However, there is still a lot to be done, including sarcasm detection, domain dependency, multilingual analysis and dealing with conflicted sentiments in a single text. In order to address these issues, this study discusses various state of the art deep learning techniques such as Convolutional Neural Networks (CN), Recurrent Neural Networks (RN), and Transformer based models, BERT, improving the performance and accuracy of ABSA. Additionally, we introduce new sentiment classification enhancement methods based on domain specific lexicon, hybrid learning, as well as latest state of the art pre-trained models. Their findings pave the way for designing more robust sentiment systems as used in product reviews, social media sentiment, etc., with an aim to move sentiment detection and interpretation further into the state-of-the-art.