International Journal of Science, Engineering and TechnologyAuthors: Pooja Dwivedi, Ashish Shukla
Abstract: With the rise in user-generated content generation on digital media, brand perception monitoring has become unfeasible using manual methods. In this paper, we present an AI-powered sentiment analysis approach that aims at reputation management of brands across social media platforms and e-commerce review channels. For this purpose, the proposed system adopts a hybrid deep learning architecture based on contextual embeddings generated by transformers and bidirectional recurrent networks for capturing sophisticated sentiments in informal text. To handle various forms of noise along with domain-specific linguistic patterns, multi-stage text preprocessing pipeline is employed. The proposed methodology uses aspect-based sentiment extraction to provide a fine-grained reputation analysis at the aspect level, not document-level sentiment polarity. The experiments conducted on a specially curated dataset of social media posts and product reviews demonstrate the ability of our framework to process real-time data streams in a consistent manner. Comparative analysis of the performance against existing state-of-the-art approaches highlights the benefits of hybrid models in dealing with sarcasm, implicit sentiment, and multilingual sentiment expressions.
DOI: https://doi.org/10.5281/zenodo.23530964
