Data Dialogue: Comparing Songs without Listening: From Mathematics, Statistics, and Computer Science to Music and Back Again

Abstract: Music is deeply entrenched in our daily lives, from our playlists to the background songs in our favorite television shows. The multidisciplinary field of MusicInformation Retrieval (MIR) is motivated by the comparisons that we, as humans, make about music and the various contexts of these comparisons. By defining tasks such as building better song recommendation systems or finding structural information in a given recording, MIR seeks to algorithmically make these musical comparisons in the same manner that a human being would, but on a much larger scale. In this talk, we will introduce the field of MIR, including popular tasks and cutting edge techniques. Then we will present aligned hierarchies, a structure-based representation that can be used for comparing songs, and new extensions of aligned hierarchies that leverage ideas from topological data analysis.