Implementing an LLM-powered recommendation system

Now that we have covered some theory about recommendation systems and emerging research on how LLMs can enhance them, let’s start building our recommendation app, which will be a movie recommender system called MovieHarbor. The goal will be to make it as general as possible, meaning that we want our app to be able to address various recommendations tasks with a conversational interface. The scenario we are going to simulate will be that of the so called “cold-start”, that means the first interaction of an user with the recommendation system, so that we do not have user’s preference history. On the other hand, we will leverage a movie database with textual description.With this purpose, we will use the “Movie recommendation data” dataset, available on Kaggle at https://www.kaggle.com/datasets/rohan4050/movie-recommendation-data.The idea is that of having a dataset with a textual description of each...