In two days (11/13!) we have the pleasure of hearing from Katie Kinnaird
who we're gotten to know over the past couple months. Her title and
abstract are below.

Same time (noon), same place (Farrell)!



Title: Hierarchical Representation for High-Dimensional Data in Music Tasks

Abstract: We propose a novel hierarchical representation for
high-dimensional, time-sequenced data that encodes relevant
size-appropriate information at several scales. We apply this
representation to Music Information Retrieval (MIR) structure tasks, by
building a signature for each digital score in our data set of Chopin
Mazurka scores. This signature captures repeated structure at several
scales and is naturally hierarchical. Given a particular MIR structure
task, such as locating a chorus or motif of a given musical song or score,
we restrict our analysis to found repetitions of a task-appropriate range
of scales. Including repetitive structure of all possible scales allows our
representation to be applicable to a more diverse set of MIR structure
tasks than other structure representations in the literature that focus
only on the coarsest level of structure.