Treating Inherent Ambiguity in Ground Truth Data: Evaluating a Chord Labelling Algorithm

Müllensiefen, Daniel; Lewis, David; Rhodes, Christophe and Wiggins, Geraint. 2007. 'Treating Inherent Ambiguity in Ground Truth Data: Evaluating a Chord Labelling Algorithm'. In: 8th International Conference on Music Information Retrieval (ISMIR). Vienna, Austria 23 - 27 September 2007. [Conference or Workshop Item]

Text (Evaluating a Chord Labelling Algorithm)
COM-Rhodes2007a.pdf - Published Version

Download (23kB) | Preview
Image (7 Steps on the Stairway to Heaven or How to evaluate algorithm for ambiguous MIR tasks)
COM-Rhodes2007b-Poster.jpg - Presentation

Download (855kB) | Preview

Abstract or Description

This paper outlines a method for evaluating a new chord-labelling algorithm using symbolic data as input. Excerpts from full-score transcriptions of 40 pop songs are used. The accuracy of the algorithm’s output is compared with that of chord labels from published song books, as assessed by experts in pop music theory. We are interested not only in the accuracy of the two sets of labels but also in the question of potential harmonic ambiguity as reflected the judges’ assessments. We focus, in this short paper, on outlining the general approach of this research project.

Item Type:

Conference or Workshop Item (Poster)

Departments, Centres and Research Units:

Computing > Intelligent Sound and Music Systems (ISMS)


September 2007Published

Event Location:

Vienna, Austria

Date range:

23 - 27 September 2007

Item ID:


Date Deposited:

26 Mar 2014 10:42

Last Modified:

21 Dec 2022 14:33


View statistics for this item...

Edit Record Edit Record (login required)