hu-music / Call-Response

Responding to the Call: Exploring Automatic Music Composition Using a Knowledge-Enhanced Model

Responding to the Call: Exploring Automatic Music Composition Using a Knowledge-Enhanced Model

This is the dataset and code implementation of AAAI2024 paper "Responding to the Call: Exploring Automatic Music Composition Using a Knowledge-Enhanced Model".

CRD Dataset Zip File

• Download CRD Dataset (19,155 Call-Response pairs in MIDI format): here

• You can also start with compound word representations after preprocessing, which is available here. It includes the training dataset, test dataset and corresponding external knowledge candidates.

Code

This folder contains the necessary codes and tutorials for running the model. We include a subset of the dataset for code demonstration.

Data Pre-Processing Script

We also provide data preprocessing scripts. It will allow you to quickly process CRD Files and your external knowledge data samples (MIDI) into the Compound Word representation as like Compound Word Transformer.

Demo

We also provide a demo of machine-composed responses here.