Content-based Original Pilipino Music (OPM) recommender system centered on mood

Abstract

Recommender systems have become increasingly relevant due to the userfocused service delivery of streaming giants like Netflix and Spotify. Systems like this make it possible to offer media content that is highly similar to a user's preference and, if available, streaming history. Despite the availability of these enterprise recommender systems, recommendations are very global and general, requiring highly technical proficiency before one can use the technology to get recommendations for a very specific genre like Original Pilipino Music (OPM). This special project proposes a web-based recommender system using content-based filtering to recommend Original Pilipino Music (OPM) to music based on their mood, with recommendations subject to the user's input (e.g. is this song happy or sad?) that will be considered in the final recommendation. This project will also leverage audio content features available from the Spotify database for the model training part of the recommender system. Both objective and subjective evaluations of the implemented recommender system are scoped down to the available audio features of available OPM or Filipino songs on the Spotify Database. While this project concurs with the findings of many researchers that music based solely on audio content or features does not provide very accurate recommendations, this project is a good minimum viable product (MVP) for music enthusiasts. Furthermore, the project is also proof of concept for a more robust OPM recommender system in the future, as detailed in the results and recommendations.