Music subject classification based on lyrics and user interpretations

Kahyun Choi, Jin Ha Lee, Xiao Hu, J. Stephen Downie

Research output: Contribution to journalArticle

Abstract

That music seekers consider song subject metadata to be helpful in their searching/browsing experience has been noted in prior published research. In an effort to develop a subject-based tagging system, we explored the creation of automatically generated song subject classifications. Our classifications were derived from two different sources of song-related text: 1) lyrics; and 2) user interpretations of lyrics collected from songmeanings.com. While both sources contain subject-related information, we found that user-generated interpretations always outperformed lyrics in terms of classification accuracy. This suggests that user interpretations are more useful in the subject classification task than lyrics because the semantically ambiguous poetic nature of lyrics tends to confuse classifiers. An examination of top-ranked terms and confusion matrices supported our contention that users' interpretations work better for detecting the meaning of songs than what is conveyed through lyrics.

Original languageEnglish (US)
Pages (from-to)1-10
Number of pages10
JournalProceedings of the Association for Information Science and Technology
Volume53
Issue number1
DOIs
StatePublished - Jan 1 2016

Fingerprint

song
music
interpretation
Metadata
Classifiers
examination
experience

Keywords

  • Interpretations of Lyrics
  • Music Digital Library
  • Music Subject Classification
  • Subject Metadata

ASJC Scopus subject areas

  • Computer Science(all)
  • Library and Information Sciences

Cite this

Music subject classification based on lyrics and user interpretations. / Choi, Kahyun; Lee, Jin Ha; Hu, Xiao; Downie, J. Stephen.

In: Proceedings of the Association for Information Science and Technology, Vol. 53, No. 1, 01.01.2016, p. 1-10.

Research output: Contribution to journalArticle

@article{914a7b236eef46caaf3203fc02f99486,
title = "Music subject classification based on lyrics and user interpretations",
abstract = "That music seekers consider song subject metadata to be helpful in their searching/browsing experience has been noted in prior published research. In an effort to develop a subject-based tagging system, we explored the creation of automatically generated song subject classifications. Our classifications were derived from two different sources of song-related text: 1) lyrics; and 2) user interpretations of lyrics collected from songmeanings.com. While both sources contain subject-related information, we found that user-generated interpretations always outperformed lyrics in terms of classification accuracy. This suggests that user interpretations are more useful in the subject classification task than lyrics because the semantically ambiguous poetic nature of lyrics tends to confuse classifiers. An examination of top-ranked terms and confusion matrices supported our contention that users' interpretations work better for detecting the meaning of songs than what is conveyed through lyrics.",
keywords = "Interpretations of Lyrics, Music Digital Library, Music Subject Classification, Subject Metadata",
author = "Kahyun Choi and Lee, {Jin Ha} and Xiao Hu and Downie, {J. Stephen}",
year = "2016",
month = "1",
day = "1",
doi = "10.1002/pra2.2016.14505301041",
language = "English (US)",
volume = "53",
pages = "1--10",
journal = "Proceedings of the Association for Information Science and Technology",
issn = "2373-9231",
publisher = "John Wiley and Sons Inc.",
number = "1",

}

TY - JOUR

T1 - Music subject classification based on lyrics and user interpretations

AU - Choi, Kahyun

AU - Lee, Jin Ha

AU - Hu, Xiao

AU - Downie, J. Stephen

PY - 2016/1/1

Y1 - 2016/1/1

N2 - That music seekers consider song subject metadata to be helpful in their searching/browsing experience has been noted in prior published research. In an effort to develop a subject-based tagging system, we explored the creation of automatically generated song subject classifications. Our classifications were derived from two different sources of song-related text: 1) lyrics; and 2) user interpretations of lyrics collected from songmeanings.com. While both sources contain subject-related information, we found that user-generated interpretations always outperformed lyrics in terms of classification accuracy. This suggests that user interpretations are more useful in the subject classification task than lyrics because the semantically ambiguous poetic nature of lyrics tends to confuse classifiers. An examination of top-ranked terms and confusion matrices supported our contention that users' interpretations work better for detecting the meaning of songs than what is conveyed through lyrics.

AB - That music seekers consider song subject metadata to be helpful in their searching/browsing experience has been noted in prior published research. In an effort to develop a subject-based tagging system, we explored the creation of automatically generated song subject classifications. Our classifications were derived from two different sources of song-related text: 1) lyrics; and 2) user interpretations of lyrics collected from songmeanings.com. While both sources contain subject-related information, we found that user-generated interpretations always outperformed lyrics in terms of classification accuracy. This suggests that user interpretations are more useful in the subject classification task than lyrics because the semantically ambiguous poetic nature of lyrics tends to confuse classifiers. An examination of top-ranked terms and confusion matrices supported our contention that users' interpretations work better for detecting the meaning of songs than what is conveyed through lyrics.

KW - Interpretations of Lyrics

KW - Music Digital Library

KW - Music Subject Classification

KW - Subject Metadata

UR - http://www.scopus.com/inward/record.url?scp=85015322338&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85015322338&partnerID=8YFLogxK

U2 - 10.1002/pra2.2016.14505301041

DO - 10.1002/pra2.2016.14505301041

M3 - Article

AN - SCOPUS:85015322338

VL - 53

SP - 1

EP - 10

JO - Proceedings of the Association for Information Science and Technology

JF - Proceedings of the Association for Information Science and Technology

SN - 2373-9231

IS - 1

ER -