InterLied: Toolkit for Computational Music Analysis

InterLied is an interface, run by a set of algorithms for music score (primarily lieder) analysis. The project is musicologically motivated, thus it streams towards wider accessibility. It not only consists of the algorithms for music (pattern) analysis, but also provides the informative jupyter notebook and a case study that serves as a tutorial for less computationally experienced users. In order to make the codes more accessible, the project also allows the user to run the algorithms through the inter-face (GUI), a convenient format for users with little computational knowledge. The latter allows the user to apply the algorithms to their corpora without direct inter-action with the code. The algorithms’ primary focus is a “music pattern” – melodic, rhythmic, and textual (e.g., lyric), but they also explore metadata information. While music parameters, such as key, meters and other musical features are being studied with music21 Python library, the patterns and lyrics are analysed with meth-ods from computational linguistics, such as n-grams and word tokenization. Aside from thorough explanation of the benefits of the algorithms, this thesis also focuses on limitations of computationally applied methods, addresses their disadvantages and problematises the current issues of digital humanities. Both, advantages and disadvantages of applied computational approach are being explored through two case studies, demonstrated on 63 Slovenian 20th century lieder (“art” songs) by three composers - Marij Kogoj, Slavko Osterc and Lucijan Marija Škerjanc, first focusing on musical features alone, and second including lyrics and metadata information.