Elamur Research⌕
☰
home / papers / 2024 / A Functional Trade-off between Prosodic and Semantic Cues in Conveying Sarcasm

A Functional Trade-off between Prosodic and Semantic Cues in Conveying Sarcasm

Paper · 2024 · Zhu Li, Xiyuan Gao, Yuqing Zhang et al.

Summary

Studies of how sarcasm sounds disagree, especially for English. Some report higher pitch, others lower, and spontaneous speech often shows few reliable markers. The authors propose that part of this inconsistency comes from mixing sarcasm types. Their hypothesis is that prosody and meaning compensate for each other: speakers lean on voice cues when the words alone do not reveal the irony, and lean on them less when the wording already makes it obvious.

They draw on MUStARD++, a corpus of sitcom clips labelled with three sarcasm types:

  • embedded: the incongruity is inside the sentence;
  • propositional: context is needed to see it;
  • illocutionary: it is carried by non-textual signals.

They measure pitch, amplitude and duration in Praat, both over whole utterances and on annotated key phrases. Each key phrase is compared against the same words spoken neutrally by the same speaker. Linear mixed-effects models show that prosodic cues matter less at the phrase level when the semantics already signal sarcasm. This supports a functional trade-off and matters for spoken dialogue systems that must detect sarcasm.

Key points

  • Data source: MUStARD++ has 1202 audiovisual utterances, half of them sarcastic. The sarcastic part splits into 333 propositional, 87 embedded and 178 illocutionary utterances. Of the illocutionary ones, 44 that were signalled only visually were dropped, leaving 134.
  • Key-phrase annotation: Two linguistics graduate students marked key phrases in Big Bang Theory sarcastic utterances. These were content words plus exclamations such as "oh" and "yeah". They then judged candidate neutral counterparts from the transcripts, reaching Kappa 0.6042. Only items both annotators called neutral were kept.
  • Final set: 114 key phrases (38 per sarcasm type), each with at least two non-sarcastic counterparts, for 361 audio samples in total.
  • Tools and measures:
    • voicefixer removed noise and laughter.
    • Gentle supplied word-level timestamps.
    • Praat extracted mean F0, F0 range, mean amplitude and amplitude range, plus duration for key phrases. Speech rate was dropped because of pauses and laughter.
    • Models were fitted with lme4 and tested with lmerTest in R.
  • Main finding: Key phrases in semantically transparent (embedded) sarcasm show weaker prosodic marking than key phrases whose sarcasm is not evident from meaning.

Interesting ideas

  • I1. Control the lexical confound by comparing the identical phrase from the same speaker in sarcastic versus neutral use, so that only the delivery differs.
  • I2. Classify sarcasm by where its evidence lives: inside the sentence, in the surrounding context, or outside the text entirely.
  • I3. Cues trade off: when one channel carries the signal strongly, the others are used less. The absence of textual incongruity therefore does not mean sincerity, because the signal may sit in a channel the system cannot see.
  • I4. Locate sarcasm on a few semantically critical anchor words (content words and interjections) rather than spreading it over the whole utterance.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (learn): Build negative controls for irony evaluation by pairing each ironic test sentence with the same wording in a literal context, and check that the irony detector flips only on the context.
  • I2 (frames): The irony component in prag could record which source licensed its reading (internal incongruity versus a clash with prior frames), keeping each judgement explainable.
  • I3 (answer): Readers answering questions about tales or dialogue should flag a literal reading as uncertain when the context suggests irony but the text itself shows no incongruity, rather than asserting sincerity.
  • I4 (frames): Irony detection could anchor on specific UD nodes, such as evaluative adjectives or nouns and discourse interjections, then test those against the world state, instead of scoring whole sentences.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Arts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksPsychology · 1,552 worksExperimental and Cognitive Psychology · 870 works

🧭 Topics 3Research subjects and languages the work is about.English · 2,263 worksSentiment & emotion · 972 worksSpeech & spoken language · 480 works

Cite

Cite as Zhu Li, Xiyuan Gao, Yuqing Zhang, Shekhar Nayak and Matt Coler. 2024. A Functional Trade-off between Prosodic and Semantic Cues in Conveying Sarcasm. arXiv:2408.14892.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.21437/interspeech.2024-1962 · arXiv 2408.14892

@misc{li2024functional,
  title = {{A Functional Trade-off between Prosodic and Semantic Cues in Conveying Sarcasm}},
  author = {Zhu Li, Xiyuan Gao, Yuqing Zhang, Shekhar Nayak and Matt Coler},
  year = {2024},
  doi = {10.21437/interspeech.2024-1962},
  eprint = {2408.14892},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.21437/interspeech.2024-1962}
}

References

22 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. D. Sperber and D. Wilson, “Irony and the use-mention distinction,” Philosophy, vol. 3, pp. 143–184, 1981. find
  2. H. Grice, “Logic and conversation,” Syntax and Semantics, vol. 3, pp. 43–58, 1975. ↗
  3. L. Scharrer and U. Christmann, “Voice modulations in german ironic speech,” Language and speech, vol. 54, no. 4, pp. 435–465, 2011. ↗
  4. P. Rockwell, “Lower, slower, louder: Vocal cues of sarcasm,” Journal of Psycholinguistic research, vol. 29, pp. 483–495, 2000. ↗
  5. H. S. Cheang and M. D. Pell, “The sound of sarcasm,” Speech communication, vol. 50, no. 5, pp. 366–381, 2008. ↗
  6. L. Anolli, R. Ciceri, and M. G. Infantino, “Irony as a game of implicitness: Acoustic profiles of ironic communication,” Journal of Psycholinguistic Research, vol. 29, pp. 275–311, 2000. ↗
  7. H. S. Cheang and M. D. Pell, “Acoustic markers of sarcasm in cantonese and english,” The Journal of the Acoustical Society of America, vol. 126, no. 3, pp. 1394–1405, 2009. ↗
  8. H. Loevenbruck, M. B. Jannet, M. d’Imperio, M. Spini, and M. Champagne-Lavau, “Prosodic cues of sarcastic speech in french: slower, higher, wider,” in Interspeech 2013-14th Annual Conference of the International Speech Communication Association, 2013, pp. 3537–3541. find
  9. A. Chen and L. Boves, “What’s in a word: Sounding sarcastic in british english,” Journal of the International Phonetic Association, vol. 48, no. 1, pp. 57–76, 2018. find
  10. G. A. Bryant and J. E. Fox Tree, “Is there an ironic tone of voice?” Language and speech, vol. 48, no. 3, pp. 257–277, 2005. ↗
  11. L. Anolli, R. Ciceri, and M. G. Infantino, “From “blame by praise” to “praise by blame”: Analysis of vocal patterns in ironic communication,” International Journal of Psychology, vol. 37, no. 5, pp. 266–276, 2002. find
  12. C. A. Fowler and J. Housum, “Talkers’ signaling of “new” and “old” words in speech and listeners’ perception and use of the distinction,” Journal of memory and language, vol. 26, no. 5, pp. 489–504, 1987. find
  13. A. Ray, S. Mishra, A. Nunna, and P. Bhattacharyya, “A multimodal corpus for emotion recognition in sarcasm,” in Proceedings of the Thirteenth Language Resources and Evaluation Conference, N. Calzolari, F. Béchet, P. Blache, K. Choukri, C. Cieri, T. Declerck, S. Goggi, H. Isahara, B. Maegaard, J. Mariani, H. Mazo, J. Odijk, and S. Piperidis, Eds. Marseille, France: European Language Resources Association, Jun. 2022, pp. 6992–7003. [Online]. Available: https://aclanthology.org/2022.lrec-1.756 ↗
  14. Z. Zvolenszky, “A gricean rearrangement of epithets,” 2012. find
  15. P. Boersma, “Praat: doing phonetics by computer [computer program],” http://www.praat. org/, 2011. ↗
  16. R Core Team, R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria, 2023. [Online]. Available: https://www.R-project.org/ ↗
  17. D. Bates, M. Mächler, B. Bolker, and S. Walker, “Fitting linear mixed-effects models using lme4,” Journal of Statistical Software, vol. 67, no. 1, pp. 1–48, 2015. ↗
  18. A. Kuznetsova, P. B. Brockhoff, and R. H. B. Christensen, “lmertest package: tests in linear mixed effects models,” Journal of statistical software, vol. 82, no. 13, 2017. ↗
  19. K. R. Scherer, D. R. Ladd, and K. E. Silverman, “Vocal cues to speaker affect: Testing two models,” The Journal of the Acoustical Society of America, vol. 76, no. 5, pp. 1346–1356, 1984. ↗
  20. D. R. Ladd, K. E. Silverman, F. Tolkmitt, G. Bergmann, and K. R. Scherer, “Evidence for the independent function of intonation contour type, voice quality, and f 0 range in signaling speaker affect,” The Journal of the Acoustical Society of America, vol. 78, no. 2, pp. 435–444, 1985. find
  21. X. Gao, S. Nayak, and M. Coler, “Deep CNN-based Inductive Transfer Learning for Sarcasm Detection in Speech,” in Proc. Interspeech 2022, 2022, pp. 2323–2327. find
  22. Z. Li, X. Gao, S. Nayak, and M. Coler, “Sarcasticspeech: Speech synthesis for sarcasm in low-resource scenarios,” in 12th ISCA Speech Synthesis Workshop (SSW2023). ISCA, 2023, pp. 242– 243. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/li-2024-functional-trade-off-between-prosodic/