Classically-trained song editors have produced many songs, including hymns, which generate guitar chords that guitarists don’t want to play. This is because there is a strong disconnect between how these classical composers view music and the way that modern guitarists view music. The guitar chords used for a song often sound bad played with piano accompaniment because they have been oversimplified to match keyboard harmonization. Some problems are that harmonies don’t match up closely, and then the piece sounds bad. Other times, the guitar arrangements are too simple or complicated to be effective. Guitars are extremely common in worship settings, and because of this, it is important to have suitable guitar chords available. Our goal is to automatically generate guitar chords for these different situations.
The solution I have provided generates guitar chords with a given score with about 75% accuracy. Accuracy was determined by a chord by chord comparison. I used a machine learning system to implement this solution. The softwares used were Music21 for data manipulation, MusicXML files for reading music scores and Neural Machine Translation(NMT) for the machine learning.
Michelle Ferdinands is a Senior Computer Science student(BCS) with a double major in Mathematics.