This repo contains the framework for a Song Recommendation Engine based on Spotify playlists using Word2Vec. This end-to-end project starts by scrapping data from Spotify's API, collecting over 200k playlists and 8.411.437 songs, creating a sqlite database to store them. After that, an off-the-shelf model Word2Vec is trained so that songs that appear on the same context (i.e. same playlists) will have similar vector representations, much like words in sentences. Finally, a web application is deployed using Streamlit, where the recommendation engine can be tested.
This project was inspired by code found in the Deep Learning Cookbook, by Douwe Osinga.
