Development of a Chatbot for Movie Recommendations Based on User Text Requests.

Due to the increased use of streaming services, the need to recommend movies through efficient systems cannot be overemphasized. This thesis aims to construct a chatbot for movie recommendation using text requests from the user and using the BERT algorithm for NLP. The URLs of the films, images, titles, and descriptions were collected from the Mail.ru film catalog and preprocessed to generate the intermediate vectors. The architecture includes a chatbot used for matching the user’s queries to the movies through embeddings and Faiss indexing for an efficient search and movie recommendations. The chatbot was developed using the Telebot for the Telegram Bot and Streamlit for the web interface. This work used qualitative survey feedback as criteria for assessment. Interacting with the chatbot, one is able to get a satisfying response and it can be concluded that the application of NLP together with a real time interaction can be used for making improvements in movie recommendations. Possibilities for the further work include the incorporation of other data concerning the user and the investigation of different algorithms to enhance the model’s precision.