Computer Science applications to MIR
DIDONE Project Seminar Series
21 January 2021, 12am-1pm ECT
Ana Llorens and Paula Muñoz-Lago
In this seminar, we will be presenting the computational methodology we are applying to the study of our musical scores in DIDONE.
First, we will discuss the general approach to data extraction and processing, with a special focus on the musical features we consider relevant to our corpus. Subsequently, and for the most part of the seminar, we will present our latest software for the analysis of melodic features in XML scores. In particular, the software retrieves and computes data on melodic configuration (non-enharmonic intervals, interval types, scale degrees, ambitus, leaps, etc.) in the form of Excel spreadsheets and provides accompanying visualisations for easier evaluation. It can process multiple files at a time and automatically groups the pieces for analysis according to select parameters (date, composer, key, etc.) and at various levels of filtering.
Finally, we will show some ongoing artificial intelligence applications on thusly-extracted data. These mainly focus on unsupervised learning for corpus clustering attending to compositional features. Although not necessarily restricted to it, such techniques will ultimately be aimed at determining whether or not there is a direct correlation between compositional strategies and emotional content in music, in line with DIDONE’s general aim.
More info, and recordings of previous ones here.