Using self-organizing maps to cluster music files based on lyrics and audio features

Date of Publication


Document Type

Master's Thesis

Degree Name

Master of Science in Computer Science


College of Computer Studies


Computer Science

Thesis Adviser

Arnulfo Azcarraga


Music is an integral part of our everyday lives, and with the advent of portable music players with large storage capacities to maintain large archives of songs, users need to be able to organize their archives in a manner that would allow them to retrieve the songs that they want to listen. It would also be useful for he archive to be able to recommend songs related to the song that has just been played, as a new way of accessing music archives. This research studies the effect of clustering music files using both audio features and song lyrics. Audio features refer to the melody, pitch and other audible features of the song, whereas song lyrics refer to the words used in the song. Clusters of these music files produced by Self-Organizing Maps (SOM) are compared with the natural groupings of the songs based on their actual genre. This research shows that using lyrics alone to cluster music files is not enough. However, when the lyrics and audio are used at the same time, or when audio is used to refine the clusters already made by lyrics, a more organized SOM can be produced.

Abstract Format






Accession Number


Shelf Location

Archives, The Learning Commons, 12F Henry Sy Sr. Hall

Physical Description

xi, 94 leaves ; 28 cm. + 1 computer optical disc.


Music; Tone clusters; Self-organizing maps

This document is currently not available here.