The Album Discoverer - An Album Recommendation System

9:35am - 10:00am on Friday, October 2 in Online

Angeline Protacio



Most music recommendation systems rely heavily on individual tracks. Track recommendations often drive playlist creation, and are often used to introduce new artists to a broader audience. However, musicians often conceptualize their work as an album, with an intended track order, and overarching themes. I wanted to create a music recommender that focused on recommending albums, rather than individual songs, to preserve this thoughtfulness and provide a more complete experience for the listener, as the artist intended. In this talk, I’ll discuss how I built the Album Discoverer, a Flask app that uses principal components analysis to distill an album down to its most salient parts, and recommends albums with the highest cosine similarity. Along the way, I will show how I validated each step of the analysis, and end with a demonstration of the app.

Want to edit this page?