Automatic Harmonization using Recurrent Neural Networks

Project information

In this work, we apply RNNs to the music field with the aim of using the Annotated Beethoven Corpus (ABC) to learn the underlying structure of the chord sequences in Beethoven’s string quartets. Our model is aimed to predict the chords that follow a given sequence of arbitrary length provided by the user and it also offers the option to condition the predicted chords by some features like the global key or the length of the phrases.