Advanced search
2 files | 2.49 MB Add to list

Joint emotion label space modeling for affect lexica

Author
Organization
Project
Abstract
Emotion lexica are commonly used resources to combat data poverty in automatic emotion detection. However, vocabulary coverage issues, differences in construction method and discrepancies in emotion framework and representation result in a heterogeneous landscape of emotion detection resources, calling for a unified approach to utilizing them. To combat this, we present an extended emotion lexicon of 30,273 unique entries, which is a result of merging eight existing emotion lexica by means of a multi-view variational autoencoder (VAE). We showed that a VAE is a valid approach for combining lexica with different label spaces into a joint emotion label space with a chosen number of dimensions, and that these dimensions are still interpretable. We tested the utility of the unified VAE lexicon by employing the lexicon values as features in an emotion detection model. We found that the VAE lexicon outperformed individual lexica, but contrary to our expectations, it did not outperform a naive concatenation of lexica, although it did contribute to the naive concatenation when added as an extra lexicon. Furthermore, using lexicon information as additional features on top of state-of-the-art language models usually resulted in a better performance than when no lexicon information was used.
Keywords
Human-Computer Interaction, Theoretical Computer Science, Software, lt3, NLP, Emotion detection, Emotion lexica, VAE, NORMS

Downloads

  • (...).pdf
    • full text (Published version)
    • |
    • UGent only
    • |
    • PDF
    • |
    • 1.44 MB
  • (...).pdf
    • full text (Accepted manuscript)
    • |
    • UGent only (changes to open access on 2023-02-01)
    • |
    • PDF
    • |
    • 1.06 MB

Citation

Please use this url to cite or link to this publication:

MLA
De Bruyne, Luna, et al. “Joint Emotion Label Space Modeling for Affect Lexica.” COMPUTER SPEECH AND LANGUAGE, vol. 71, 2022, doi:10.1016/j.csl.2021.101257.
APA
De Bruyne, L., Atanasova, P., & Augenstein, I. (2022). Joint emotion label space modeling for affect lexica. COMPUTER SPEECH AND LANGUAGE, 71. https://doi.org/10.1016/j.csl.2021.101257
Chicago author-date
De Bruyne, Luna, Pepa Atanasova, and Isabelle Augenstein. 2022. “Joint Emotion Label Space Modeling for Affect Lexica.” COMPUTER SPEECH AND LANGUAGE 71. https://doi.org/10.1016/j.csl.2021.101257.
Chicago author-date (all authors)
De Bruyne, Luna, Pepa Atanasova, and Isabelle Augenstein. 2022. “Joint Emotion Label Space Modeling for Affect Lexica.” COMPUTER SPEECH AND LANGUAGE 71. doi:10.1016/j.csl.2021.101257.
Vancouver
1.
De Bruyne L, Atanasova P, Augenstein I. Joint emotion label space modeling for affect lexica. COMPUTER SPEECH AND LANGUAGE. 2022;71.
IEEE
[1]
L. De Bruyne, P. Atanasova, and I. Augenstein, “Joint emotion label space modeling for affect lexica,” COMPUTER SPEECH AND LANGUAGE, vol. 71, 2022.
@article{8730709,
  abstract     = {{Emotion lexica are commonly used resources to combat data poverty in automatic emotion detection. However, vocabulary coverage issues, differences in construction method and discrepancies in emotion framework and representation result in a heterogeneous landscape of emotion detection resources, calling for a unified approach to utilizing them. To combat this, we present an extended emotion lexicon of 30,273 unique entries, which is a result of merging eight existing emotion lexica by means of a multi-view variational autoencoder (VAE). We showed that a VAE is a valid approach for combining lexica with different label spaces into a joint emotion label space with a chosen number of dimensions, and that these dimensions are still interpretable. We tested the utility of the unified VAE lexicon by employing the lexicon values as features in an emotion detection model. We found that the VAE lexicon outperformed individual lexica, but contrary to our expectations, it did not outperform a naive concatenation of lexica, although it did contribute to the naive concatenation when added as an extra lexicon. Furthermore, using lexicon information as additional features on top of state-of-the-art language models usually resulted in a better performance than when no lexicon information was used.}},
  articleno    = {{101257}},
  author       = {{De Bruyne, Luna and Atanasova, Pepa and Augenstein, Isabelle}},
  issn         = {{0885-2308}},
  journal      = {{COMPUTER SPEECH AND LANGUAGE}},
  keywords     = {{Human-Computer Interaction,Theoretical Computer Science,Software,lt3,NLP,Emotion detection,Emotion lexica,VAE,NORMS}},
  language     = {{eng}},
  pages        = {{20}},
  title        = {{Joint emotion label space modeling for affect lexica}},
  url          = {{http://dx.doi.org/10.1016/j.csl.2021.101257}},
  volume       = {{71}},
  year         = {{2022}},
}

Altmetric
View in Altmetric
Web of Science
Times cited: