This document discusses and compares different topic modeling techniques for a movie recommendation system. It aims to recommend movies based on plot summaries without any user preference data. It tests Latent Semantic Analysis (LSA) and Latent Dirichlet Allocation (LDA) on movie plot vectors from IMDb databases. The results show that LSA performs better by recommending movies from the same genre or saga, while LDA recommends unrelated movies. LSA is better able to capture the themes in movie plots for generating relevant movie recommendations.