‘Texting Scarlatti’ is the first comprehensive study of Domenico Scarlatti’s keyboard sonatas based on analysis of over 3,300 contemporary manuscripts and printed copies. This pioneering work offers a new model for musicological research, combining traditional textual scholarship with cutting-edge technology and collaborative crowdsourcing methods.
What we know of the relationships between the surviving Scarlatti keyboard sources has changed little since the American musicologist Joel Sheveloff's landmark 1970 dissertation. While new sources have come to light, placing them within their context has remained a monumental task: the number of extant sources and variant readings mean that traditional stemmatic analysis, as Sheveloff attempted, is completely impractical.
The project team have compared almost all surviving eighteenth-century sources, creating an unprecedented dataset of 190,000 variants recorded in a searchable format readable both by humans and by computers. As a DISKAH Fellow, Dr van der Klis explores how computational phylogenetic analysis – successfully applied to manuscript traditions in literature and recently to musical sources by Windram et al. (2014; 2022) – can aid experts in their analysis of the witnesses. Findings to date have revealed unexplored links between sources suggesting previously unknown routes of transmission. At scale this approach will transform our understanding of the eighteenth-century manuscript transmission of Scarlatti’s keyboard music.

Skills and technical support requirements:

The DISKAH fellowship has enabled the Fellow to attend training and allocate time to progress with their project. This includes expanding knowledge and skills to finalise dataset preparation, as well as code cleaning, debugging, restructuring and employment of version control to track and manage changes. Moreover, during the fellowship, software workflows to generate structured data for the phylogenetic analytical software have been improved. This has streamlined the project’s workflow, boosting preparedness for large-scale analysis. Close collaboration with the RSE team from the Bede N8 Supercomputer enabled the Fellow to carry out individual experimentation with the DRI and has been critical to support efficient code parallelisation, installations and running on HPC to analyse datasets using different configurations.

User roles and DRI requirements:

The project’s dataset, workflow and performed analyses not only justify the need for powerful compute in terms of faster deployment but further enable shedding light to research questions in an efficient and holistic manner; something which would not be possible to explore before.

This fellowship illustrates various DRI related user-roles undertaken as part of the user journey.

As a data-generator, the Fellow was required to create a suitable dataset for the workflow. This dataset involved the digitised version of 3,300 contemporary manuscripts and printed copies of Domenico Scarlatti’s keyboard sonatas. The digitisation was performed by [...] with a quality of [x resolution]. The copyright of these datasets belongs to [...]. The infrastructure for storing the dataset is on-the-cloud storage (Google Drive) and was used to provide volunteers with access to the data.

Supported by volunteers, the Fellow also took the role of data-shaper by performing transcription into a suitable notation, and cleaning assisted by volunteers who have received training. Quality was ensured by the researcher so that this was ready for processing. The infrastructure used for this step involved on-the-cloud storage (Google Drive) and Python scripts which were deployed locally on the researcher’s PC.

Thereafter, as a computational critical interpreter the fellow used phylogenetic software to analyse the complete dataset using the Bede Tier 2 HPC DRI. While this would be an impractical and very low processing task to perform on a desktop computer, especially when experimenting with different types of configurations, the analysis of the data on Bede required less than 24 hours of runtime. The interpretation of the results revealed different trends to support stemmatic analyses, revealing the genealogical relationship between different sources of Scarlatti’s sonatas.

As a scholarly communicator, the fellow has produced a peer reviewed scholarly output describing the workflow undertaken and research insights. This output, when published, will become available through the Fellow’s scholarly profile.

Highlights and future work:

The project revealed that phylogenetic analysis of the Texting Scarlatti large dataset with the employment of HPC enhances digital musicology research in ways which were not accessible before. Enabling comparative analyses of the various configurations with different variant weighting can have a significant effect on stemmatic music analyses to reconstruct a composer’s original intent and accuracy of piece of work. As such, large-scale computational methods along with traditional musicology approaches have the potential to equip the community with powerful analytical tools and reshape musicology research.

Future work for the research includes sonata-specific or collection specific investigations and application of digital research methods in different music corpora.