RoBERTa-BiLSTM: A Context-Aware Hybrid Model for Sentiment Analysis

This project presents a context-aware sentiment analysis framework utilizing three powerful models—BiLSTM, RoBERTa-BiLSTM, and DistilBERT+GRU—on two benchmark datasets: IMDb and Twitter. Each model is trained to classify user-submitted text as positive or negative by effectively capturing contextual and semantic nuances. To enhance interpretability, we integrate Explainable AI techniques using LIME (Local Interpretable Model-agnostic Explanations), providing users with a visual explanation of the model’s decision-making process. The system features a user-friendly web interface developed using HTML, CSS, and JavaScript. Authenticated users can log in, submit input text, and receive real-time predictions along with LIME-based explanations. The integration of advanced NLP models with explainability and intuitive UI bridges the gap between deep learning and end-user trust, making this tool highly useful for applications in social media monitoring, review analysis, and public sentiment tracking.

Keywords:Sentiment Analysis, RoBERTa-BiLSTM, BiLSTM, DistilBERT+GRU, LIME, Explainable AI, IMDb Dataset, Twitter Dataset, NLP, Context-Aware Models