A Multi-Task Learning Framework for Sentiment Classification and Aspect-Based Analysis of IMDb Movie Reviews

The expanding number of user-generated information on platforms like IMDb gives a chance to examine feelings and specific characteristics in movie reviews, providing insights into public opinion on numerous film attributes. Traditional approaches frequently regard sentiment analysis and aspect extraction as separate processes, limiting the amount of insight that may be accomplished. This paper offers a multi-task learning architecture that incorporates both sentiment classification and aspect-based analysis into a single model, boosting the model’s potential to understand complicated language patterns and the dependency between sentiment and aspects. Leveraging deep learning techniques, particularly BERT (Bidirectional Encoder Representations from Transformers), the proposed framework simultaneously classifies sentiment and identifies aspect-specific sentiments within IMDb reviews, addressing the subtleties and contextdependent nature of movie critique. Experimental results suggest that our multi-task technique outperforms standard, single-task models, enabling a more detailed analysis of attitudes related with multiple factors such as acting, plot, and direction. This research contributes to the improvement of sentiment analysis approaches, delivering a comprehensive solution for processing complicated, real-world review datasets.