READ ME File For 'Chordify User Annotator Subjectivity Dataset'

Dataset DOI: 10.5258/SOTON/D1763

ReadMe Author: Anna Selway, University of Southampton

This dataset supports the thesis entitled ‘A Formula for Music Similarity: Utilising music-theoretical approaches in audible perceptions of harmonic similarity’
AWARDED BY: University of Southampton
DATE OF AWARD: 2021

DESCRIPTION OF THE DATA

This dataset is a sub-set of user data from the Chordify edit feature introduced in 2015. Due to the sheer number of edits created (and the accompanying textual data), Chordify reduced this dataset (for their research) to 11,638 edits. They did so by through filtering based on three rules: (1) the edits had to be created on a recent algorithmic output, to ensure the quality of the original labels the users had to edit; (2) the users had to have edited at least ten beats to ensure that they were actively editing the songs, not just trying out the feature; and (3) the annotators had to have edited multiple songs to ensure they had some basic experience in transcribing harmony. The outcome of this filtering resulted in the 11,638--edit subset.

This dataset was reduced further for the purpose of my thesis' research to produce a dataset suitable for a Riemannian function analysis, and of a meaningful, but manageable size for the analytical processes that needed completion by hand. The rules stipulated were that the edit had to: 
- Have a working YouTube URL.
- Have edits that span the whole song.
- Have edits that were more complex than just changing the timing of a chord.
- Have edits that were more complex than just adding sustained notes.
- Be from a song longer than 160 beats.
- Be from a song that had edits made by at least three users (and the original annotation).

In this repository you will find a total of 148 edits, from 41 different songs by 77 annotators.


This dataset contains:

These are grouped under the following locations:
1. 'original annotations': include the raw annotations made by the Chordify users. 
2. 'merged annotations': the annotations of each user aligned with each other for an individual song. 
3. 'editlinks': the metadata link to the Youtube video that these annotations were made in relation to using Chordify. 
4. 'Excel Sheet's: my personal analysis of each songs various annotations, and an overall analysis document. 


Date of data collection: Data collected between 2016 and 2018.

Information about geographic location of data collection: 

Licence:
Open

Related projects/Funders:
Funded by the EPSRC through the Web Science Doctoral Training College. 
Supported by Chordify. 

Related publication:


Date that the file was created: March, 2021

--------------