A Fine-Tuned Personalized Movie Recommendation System Using Mask Language Models

Recommendation systems have become indispensable for personalizing user experiences across various digital platforms, particularly in online entertainment. Conventional methods, such as collaborative filtering and content-based filtering, have provided basic techniques for constructing such systems; however, these methods tend to be plagued by shortcomings in semantic comprehension, cold-start issues, and the inability to adapt to sophisticated user preferences. This work introduces a novel methodology that leverages the strengths of a fine-tuned masked language model (MLM), specifically BERT, to create a personalized movie recommendation system. The system is implemented through an interactive web app developed with Streamlit, allowing users to enter their preferences, view suggested movies with preview posters, and provide feedback, which is logged persistently for future model optimization. Our analysis, conducted using the TMDB 50000 dataset and user reviews, reveals that the MLM-based method achieves higher precision and recall than conventional content-based and collaborative filtering baselines.