The website about the paper Basso Continuo Goes Digital: Collecting and Aligning a Symbolic Dataset of Continuo Performance presented at The Sixth Conference on AI Music Creativity, taking place o 10 - 12 September 2025 in Brussels, Belgium.
Adam Štefunko1, Suhit Chiruthapudi2, Jan Hajič jr.1, Carlos Eduardo Cancino-Chacón2
1Charles University, Faculty of Mathematics and Physics, Insitute of Formal and Applied Linguistics
2Johannes Kepler University, Institute of Computational Perception
The dataset is a collection of MIDI recordings obtained from 7 professional harpsichord players and university-level students. Basic demographic data about each participant can be seen in the following table:
| ID | Experience Time (years) | Practice Rate | Performance Rate | Self-Assessed Level (%) | Level |
|---|---|---|---|---|---|
| 1809 | 7 | Sporadically | A few times a month | 80 | Professional |
| 4045 | 35 | Sporadically | Every day | 100 | Professional |
| 6974 | 4 | Sporadically | Once a week | 50 | Student |
| 8352 | 1 | A few times a week | A few times a month | 35 | Student |
| 8627 | 2 | Once a week | Sporadically | 10 | Student |
| 8780 | 6 | Sporadically | A few times a week | 90 | Professional |
| 9404 | 10 | Sporadically | Every day | 75 | Professional |
You can find the dataset of 175 MIDI recordings of basso continuo performance here. The dataset directory is structured as follows:
Performances directory containing performances of 7 performers, each having their own subdirectory. Each of these subdirectories contains MIDI files for 5 performances of 5 pieces with basso continuo. The naming structure of the files is [performer_id]_[piece_id]_cembalo_[take_no].mid.Scores directory containing MusicXML, MIDI and PDF versions of five scores. There is a separate directory for each file format.Manual_alignments directory containst two subdirectories: each for just bass alignemnt ground truth manual alignments and one for the full realization alignment ground truth. Each of these subdirectories has the same structure: subdirectories named [performer_id]_[piece_id]_[take_no], each of them containing a CSV file named manual_alignment.csv.We created an alignment system, based on existing SoTA performance-to-score alignment methods, to automatically align performances from the dataset and test these different methods used. More information on the system and the code we used can be found in the system’s own repository.
Paper: https://zenodo.org/records/16946799
Software: https://github.com/adamcho14/BassoContinuoPerformanceAnalysisTools
Dataset: http://hdl.handle.net/11234/1-5963