Irish traditional music, like all vernacular artforms, has a complex and ancient root structure. Musicians, musicologists and archivists have travelled along the branches to establish the place of individual tunes in what are known as ‘tune families’.
Tune families – such as The Road to Lisdoonvarna family – contain groupings of related tunes that use similar chord patterns and are likely to be variations that drift and mutate from an original air. The Road to Lisdoonvarna is known to be part of a family which includes tunes such as The Kerry Cow, All the Way to Galway, Sarsfield’s March, Yankee Doodle and No Surrender.
What is more challenging is making connections across these families in a way that allows for searching and matching on a wider scale.
A community resource called thesession.org features over 40,000 traditional Irish tunes. An Insight team at University of Galway led by Dr James McDermott has used the resource to teach an algorithm to spot patterns in the tunes and to find previously unmade connections. The research is part of the EU Horizon 2020 Polifonia project to recreate connections between music, people and places from the 16th century to the modern day.
By analysing the tune shapes, the algorithm can find long lost relatives among these 40,000 tunes and reunite them.
The algorithm was created as part of the MSc research of traditional musician Danny Diamond, supervised by McDermott and Prof Mathieu d’Aquin (LORIA, France). As an expert musician, Diamond was able to inject knowledge of the domain into the algorithm, and to verify its findings in a curated set of known tune families.
While the algorithm will never be as adept at matching as the human ear, it can throw up potential matches that musicologists can use as a basis for further exploration. A human researcher/musician can only handle tunes in the hundreds. An AI assistant can search at scale and together they can build a more detailed history of the Irish musical tradition and allow modern players to compare repertoires and make living connections in music.
Project name: Automatic tune family detection in a corpus of Irish traditional dance tunes