---
title: "Phonetic Cues Improve LLM-Based Pun Detection in Short Text"
description: "Drawing on the General Theory of Verbal Humor, the authors build a hybrid pipeline for English short texts. 0.732 accuracy. 16 references."
authors: ["Adith Santosh Thaniserikaran", "Govind Harikrishnan"]
year: 2026
venue: "Proceedings of the 2nd Workshop on Computational Humor (CHum 2026)"
source: "https://aclanthology.org/2026.chum-1.6/"
doi: "10.18653/v1/2026.chum-1.6"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/thaniserikaran-2026-phonetic-cues-improve-llm-based/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/thaniserikaran-2026-phonetic-cues-improve-llm-based/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/39/) / Phonetic Cues Improve LLM-Based Pun Detection in Short Text

# Phonetic Cues Improve LLM-Based Pun Detection in Short Text

**Authors:** Adith Santosh Thaniserikaran and Govind Harikrishnan · **Year:** 2026 · **Venue:** [Proceedings of the 2nd Workshop on Computational Humor (CHum 2026)](https://md.research.elamur.ai/free/en/venues/chum/) ([Workshop on Computational Humor 2026](https://md.research.elamur.ai/free/en/venues/chum/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.chum-1.6/) · [PDF](https://aclanthology.org/2026.chum-1.6.pdf) · [DOI 10.18653/v1/2026.chum-1.6](https://doi.org/10.18653/v1/2026.chum-1.6) · [OpenAlex](https://openalex.org/W7166873390)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
The paper addresses a gap in LLM-based humor detection: models often sense that a short text is playful but fail to identify the ambiguity behind a pun, especially when it depends on sound rather than spelling. Drawing on the General Theory of Verbal Humor, the authors build a hybrid pipeline for English short texts. NLTK preprocessing keeps lemmatized, POS-tagged content words. Their ARPABET pronunciations come from the CMU Pronouncing Dictionary (CMUdict). A coarse weighted Levenshtein distance then flags near-homophone pairs, with cheaper substitutions within the vowel or consonant class. Strong phonetic matches are inserted into the GPT-5.4 prompt. Otherwise, GPT-5.4 searches for homograph ambiguity. Finally, GPT-5.4 labels the text Joke or Non-joke and explains its decision. On a 500-example evaluation set, the full system modestly beats a plain GPT baseline. Ablations show that most of the gain comes from phonetic cues. This suggests that symbolic sound information gives LLMs useful, interpretable evidence for sound-based wordplay.

## Key points
- **Results:** The full system reaches 0.732 accuracy and 0.720 Macro F1, compared with 0.710 and 0.692 for plain GPT-5.4.
- **Phonetic cues matter most:** Removing them (no_phonetic) causes the largest drop, to 0.610 accuracy and 0.551 Macro F1. Removing homograph detection barely changes results (0.732 / 0.718).
- **Phonetics alone fails:** The phonetic_only variant collapses to 0.504 accuracy and 0.075 Joke F1, so it is not a standalone classifier.
- **Bias toward jokes:** The system still over-predicts the Joke class. Its Non-joke F1 is 0.660, while its Joke F1 is 0.780. Witty or ambiguous non-jokes are often misclassified.
- **Limitations:** Phonetic matching works only on single tokens, so it misses multi-word homophones such as "I scream" versus "ice cream." It also approximates sound from text via CMUdict, without real audio or prosody.

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

## Institutions

- [Purdue University West Lafayette](https://md.research.elamur.ai/free/en/universities/purdue-university-west-lafayette/) — University, West Lafayette, United States · 23 works · 1,417 citations

📚 Shelves (6): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works · [💭 Psychology](https://md.research.elamur.ai/free/en/shelves/psychology/) · 1,552 works · [💭 Social Psychology](https://md.research.elamur.ai/free/en/shelves/psychology/social-psychology/) · 306 works

## Cite

Cite as: Adith Santosh Thaniserikaran and Govind Harikrishnan. 2026. Phonetic Cues Improve LLM-Based Pun Detection in Short Text. Proceedings of the 2nd Workshop on Computational Humor (CHum 2026).

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{thaniserikaran2026phonetic,
  title = {{Phonetic Cues Improve LLM-Based Pun Detection in Short Text}},
  author = {Adith Santosh Thaniserikaran and Govind Harikrishnan},
  booktitle = {Proceedings of the 2nd Workshop on Computational Humor (CHum 2026)},
  year = {2026},
  doi = {10.18653/v1/2026.chum-1.6},
  url = {https://doi.org/10.18653/v1/2026.chum-1.6}
}
```

- [← previous PHMartialLawNER: A Tagalog Named Entity Recognition Corpus for the Philippine Martial Law Era](https://md.research.elamur.ai/free/en/papers/tabuzo-2026-phmartiallawner-tagalog-named-entity-recognition/)
- [next → Phonetic-based Ranking for Improved Pseudo-Labeling in Low-Resource ASR](https://md.research.elamur.ai/free/en/papers/matassoni-2026-phonetic-based-ranking-improved-pseudo/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/39/)

## References

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

1. Salvatore Attardo, Christian F. Hempelmann, and Sara Di Maio. 2002. Script oppositions and logical mechanisms: Modeling incongruities and their resolutions. Humor, 15(1):3–46. [find](https://scholar.google.com/scholar?q=Salvatore+Attardo%2C+Christian+F%2E+Hempelmann%2C+and+Sara+Di+Maio%2E+2002%2E+Script+oppositions+and+logical+mechanisms%3A+Modeling+incongruities+and+their+resolutions%2E+Humor%2C+15%281%29%3A3%E2%80%9346%2E)
2. Adrianna Romanowski, Pedro H. V. Valois, and Kazuhiro Fukui. 2025. From punchlines to predictions: A metric to assess llm performance in identifying humor in stand-up comedy. In Proceedings of the Workshop on Cognitive Modeling and Computational Linguistics, pages 36–46. Association for Computationa… [find](https://scholar.google.com/scholar?q=Adrianna+Romanowski%2C+Pedro+H%2E+V%2E+Valois%2C+and+Kazuhiro+Fukui%2E+2025%2E+From+punchlines+to+predictions%3A+A+metric+to+assess+llm+performance+in+identifying+humor+in+stand%2Dup+comedy%2E+In+Proceedings+of+the+Wor)
3. Salvatore Attardo and Victor Raskin. 1991. Script theory revis(it)ed: joke similarity and joke representation model. Humor, 4(3–4):293–347. [find](https://scholar.google.com/scholar?q=Salvatore+Attardo+and+Victor+Raskin%2E+1991%2E+Script+theory+revis%28it%29ed%3A+joke+similarity+and+joke+representation+model%2E+Humor%2C+4%283%E2%80%934%29%3A293%E2%80%93347%2E)
4. Carnegie Mellon University. 1998. The CMU Pronouncing Dictionary. [http://www.speech.cs.cmu.edu/](http://www.speech.cs.cmu.edu/) cgi-bin/cmudict. Accessed 2026-05-12. [↗](http://www.speech.cs.cmu.edu/)
5. Yuchen Su, Shaoxin Zhong, Yonghua Zhu, Ruofan Wang, Zijian Huang, Qiqi Wang, Na Zhao, Diana Benavides-Prado, and Michael Witbrock. 2026. Words at play: Benchmarking audio pun understanding in large audio-language models. arXiv preprint arXiv:2603.18678. — [arXiv:2603.18678](https://arxiv.org/abs/2603.18678) [find](https://scholar.google.com/scholar?q=Yuchen+Su%2C+Shaoxin+Zhong%2C+Yonghua+Zhu%2C+Ruofan+Wang%2C+Zijian+Huang%2C+Qiqi+Wang%2C+Na+Zhao%2C+Diana+Benavides%2DPrado%2C+and+Michael+Witbrock%2E+2026%2E+Words+at+play%3A+Benchmarking+audio+pun+understanding+in+large+au)
6. Alessio Cocchieri, Luca Ragazzi, Paolo Italiani, Giuseppe Tagliavini, and Gianluca Moro. 2025. “What do you call a dog that is incontrovertibly true? Dogma”: Testing LLM Generalization through Humor. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1… [find](https://scholar.google.com/scholar?q=Alessio+Cocchieri%2C+Luca+Ragazzi%2C+Paolo+Italiani%2C+Giuseppe+Tagliavini%2C+and+Gianluca+Moro%2E+2025%2E+%E2%80%9CWhat+do+you+call+a+dog+that+is+incontrovertibly+true%3F+Dogma%E2%80%9D%3A+Testing+LLM+Generalization+through+Humor%2E+)
7. Kuan Lok Zhou, Jiayi Chen, Siddharth Suresh, Reuben Narad, Timothy T. Rogers, Lalit K. Jain, Robert D. Nowak, Bob Mankoff, and Jifan Zhang. 2025. Bridging the creativity understanding gap: Small-scale human alignment enables expert-level humor ranking in llms. arXiv preprint arXiv:2502.20356. [find](https://scholar.google.com/scholar?q=Kuan+Lok+Zhou%2C+Jiayi+Chen%2C+Siddharth+Suresh%2C+Reuben+Narad%2C+Timothy+T%2E+Rogers%2C+Lalit+K%2E+Jain%2C+Robert+D%2E+Nowak%2C+Bob+Mankoff%2C+and+Jifan+Zhang%2E+2025%2E+Bridging+the+creativity+understanding+gap%3A+Small%2Dscale)
8. Lionel Fontan, Isabelle Ferrane, Jérôme Farinas, Julien Pinquier, and Xavier Aumont. 2016. Using phonologically weighted levenshtein distances for the prediction of microscopic intelligibility. In Proceedings of Interspeech 2016, pages 650–654, San Francisco, USA. [find](https://scholar.google.com/scholar?q=Lionel+Fontan%2C+Isabelle+Ferrane%2C+J%C3%A9r%C3%B4me+Farinas%2C+Julien+Pinquier%2C+and+Xavier+Aumont%2E+2016%2E+Using+phonologically+weighted+levenshtein+distances+for+the+prediction+of+microscopic+intelligibility%2E+In+Pro)
9. Fabricio Goes, Piotr Sawicki, Marek Grzes, Dan Brown, and Marco Volpe. 2023. Is gpt-4 good enough to evaluate jokes? Unpublished manuscript. Manuscript. [find](https://scholar.google.com/scholar?q=Fabricio+Goes%2C+Piotr+Sawicki%2C+Marek+Grzes%2C+Dan+Brown%2C+and+Marco+Volpe%2E+2023%2E+Is+gpt%2D4+good+enough+to+evaluate+jokes%3F+Unpublished+manuscript%2E+Manuscript%2E)
10. Christian F. Hempelmann and Salvatore Attardo. 2011. Resolutions and their incongruities: Further thoughts on logical mechanisms. Humor, 24(2):125–149. [find](https://scholar.google.com/scholar?q=Christian+F%2E+Hempelmann+and+Salvatore+Attardo%2E+2011%2E+Resolutions+and+their+incongruities%3A+Further+thoughts+on+logical+mechanisms%2E+Humor%2C+24%282%29%3A125%E2%80%93149%2E)
11. Grzegorz Kondrak. 2000. A new algorithm for the alignment of phonetic sequences. In Proceedings of the 1st Meeting of the North American Chapter of the Association for Computational Linguistics, pages 288–295. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Grzegorz+Kondrak%2E+2000%2E+A+new+algorithm+for+the+alignment+of+phonetic+sequences%2E+In+Proceedings+of+the+1st+Meeting+of+the+North+American+Chapter+of+the+Association+for+Computational+Linguistics%2C+pages)
12. Vladimir I. Levenshtein. 1966. Binary codes capable of correcting deletions, insertions, and reversals. Soviet Physics Doklady, 10(8):707–710. [find](https://scholar.google.com/scholar?q=Vladimir+I%2E+Levenshtein%2E+1966%2E+Binary+codes+capable+of+correcting+deletions%2C+insertions%2C+and+reversals%2E+Soviet+Physics+Doklady%2C+10%288%29%3A707%E2%80%93710%2E)
13. Edward Loper and Steven Bird. 2002. NLTK: The Natural Language Toolkit. arXiv preprint cs/0205028. [find](https://scholar.google.com/scholar?q=Edward+Loper+and+Steven+Bird%2E+2002%2E+NLTK%3A+The+Natural+Language+Toolkit%2E+arXiv+preprint+cs%2F0205028%2E)
14. George A. Miller. 1995. [Wordnet: A lexical database for english](https://md.research.elamur.ai/free/en/papers/miller-1994-wordnet/). Communications of the ACM, 38(11):39– 41.
15. Tristan Miller, Christian Hempelmann, and Iryna Gurevych. 2017. [Semeval-2017 task 7: Detection and interpretation of english puns](https://md.research.elamur.ai/free/en/papers/miller-2017-semeval-2017-task-7-detection/). In Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017), pages 58–68, Vancouver, Canada. Association for Computational Linguistics.
16. OpenAI. 2026. Gpt-5.4. [https://developers.openai.com/api/docs/models/gpt-5.4](https://developers.openai.com/api/docs/models/gpt-5.4). Accessed 2026-05-12. [↗](https://developers.openai.com/api/docs/models/gpt-5.4)

*Updated 1 November 2026*