Text-based Movie Recommendation Web Application with DistilBERT

This project aims to develop a website for recommending movies to users, enabling them to find movies that match their preferences and reduce search time efficiently. The project began with defining the system's scope and requirements, designing the user experience (UX) and user interface (UI) on the website, and developing the movie recommendation system to optimize its performance. The data set used for training the system was collected from Kaggle, consisting of 45,000 movies released before July 2017. The recommendation system employs DistilBERT combined with Cosine Similarity and K-Nearest Neighbor (KNN) techniques to analyze keywords and find similar movies. The results are movie recommendations provided by the system, with users expressing satisfaction with both the website and the recommendation system. DistilBERT reduced the inference time from 5.5 minutes to 3.3 minutes without compromising the recommendation quality. The project successfully achieved its objectives.