Music Harmonization Using the T5 Model

Abstract

The goal of this thesis is to investigate the ability for a machine learning model to fill in the blanks of missing musical harmonic information. The model is based on T5’s architecture and is trained on specially encoded musical data in order to determine what chords would best accompany the inputted melody. This allows for a model that can provide both harmonizations and reharmonizations of a melody, both common tasks in music.