A Corpus-Based Study on Ragtime Syncopation

This paper presents a corpus-based study on syncopation patterns in ragtime. We discuss open questions on the ragtime genre and the potential of computational tools in addressing these questions, contributing to the fields of Musicology and Music Information Retrieval (MIR), and giving back to the ragtime enthusiasts community. We introduce the RAG-collection of around 11000 ragtime MIDI files collected, organised, and distributed by many ragtime lovers around the world. The collection is accompanied by a compendium, providing useful metadata on ragtime compositions. Using this collection and the compendium, we investigate syncopation patterns in ragtime melodies, for which we tailored a melody extraction algorithm. We test and confirm musicological hypotheses about the occurrence of syncopation patterns that are considered typical for ragtime on the extracted melodies. Thus, the paper presents a first step towards modelling typical characteristics of the ragtime genre, which is an important means for enabling automatic genre classification.