Elamur Research⌕
☰
home / papers / 2019 / Joint Detection and Location of English Puns

Joint Detection and Location of English Puns

Paper · 2019 · Yanyan Zou and Wei Lu

Summary

Zou and Lu address two related tasks on English puns. The first is deciding whether a short text contains a pun. The second is identifying which word carries the double meaning. Earlier systems usually treated these as separate steps. Here both are cast as one sequence labeling problem, where every word gets a tag and a text counts as punning if any word is tagged as the pun. Because each benchmark context holds at most one pun, the authors design a BPA scheme (before / pun / after) that makes this constraint structural. The model is a BiLSTM-CRF that takes three inputs: character-level features passed through highway layers, frozen GloVe vectors, and a binary flag saying which half of the sentence a word falls in. On the homographic and heterographic datasets of Miller et al. (2017), it sets new best detection F1 scores. Its location scores match or beat earlier systems. It also handles both pun types, while some prior methods covered only homographic puns.

Key points

  • Datasets (Miller et al., 2017):
    • homographic: 2,250 contexts, 1,607 containing a pun;
    • heterographic: 1,780 contexts, 1,271 containing a pun.
    • There is no official split, so the authors used 10-fold cross validation with predictions pooled across folds, following Cai et al. (2018).
  • Two tagging schemes are compared:
    • NP marks each word as pun or not pun.
    • BPA marks words before, at, and after the pun, so a text that contains no pun is tagged all B.
  • Results with the BPA scheme:
    • Detection F1 is 91.04 (homographic) and 89.60 (heterographic). This beats the NP scheme (87.69 / 87.91) and a feature-based CRF baseline (78.00 / 79.17).
    • Location F1 is 80.19 (homographic) and 79.40 (heterographic).
  • The authors also ran a location-only model trained on pun sentences, which amounts to a pipeline with perfect detection. It reached only 67.70 / 68.84 location F1. They take this as evidence that the joint setup lets detection and location help each other.
  • Puns sit in the second half of the sentence about 88% (homographic) and 92% (heterographic) of the time.
  • In the error analysis:
    • roughly 40% of errors come from words unseen in training;
    • about 40% come from failed detection;
    • very short texts are another weak spot.
  • Code is released at github.com/zoezou2015/PunLocation.

Interesting ideas

  • I1. Turning a sentence-level yes/no decision plus a word choice into one tag sequence lets a single model answer both questions. The tag alphabet (B/P/A) is shaped so that the "at most one target" rule follows from the label transitions instead of needing a separate post-check.
  • I2. Corpus statistics about where the phenomenon usually occurs are fed in as a cheap feature: a single binary first-half/second-half flag rather than a learned position embedding.
  • I3. The joint model is compared against a location-only system that is handed perfect detection. Showing that the joint system still wins makes the case for joint learning convincingly.
  • I4. The error analysis sorts mistakes into a few named buckets (unseen words, detection failures, short texts) with rough shares, which points directly to the next fix.

For Mova

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

  • I1 (frames): The prag irony/wordplay detector could use a before/target/after labeling over tree tokens to flag a sentence as punning and name the ambiguous word at once, which gives an explainable pointer to the trigger.
  • I2 (tokens): The tagger could add a coarse position-in-sentence flag where a phenomenon is known to cluster, such as sentence-final punchlines or discourse markers, while keeping the feature human-readable.
  • I3 (learn): Evaluations could add oracle-upstream baselines (e.g., gold UD trees feeding frames) to measure whether a joint or downstream stage truly benefits from the coupling or just inherits upstream quality.
  • I4 (learn): Self-induced repair rules could be prioritized by tagging each failure with a cause category and tracking bucket proportions across training rounds.

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).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 1Research subjects and languages the work is about.English · 2,263 works

Cite

Cite as Yanyan Zou and Wei Lu. 2019. Joint Detection and Location of English Puns. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/n19-1217

@inproceedings{zou2019joint,
  title = {{Joint Detection and Location of English Puns}},
  author = {Yanyan Zou and Wei Lu},
  booktitle = {Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers)},
  year = {2019},
  doi = {10.18653/v1/n19-1217},
  url = {https://doi.org/10.18653/v1/n19-1217}
}

References

32 references, 12 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Léon Bottou. 1991. Stochastic gradient learning in neural networks. In Proc. of Neuro-Nimes. find
  2. Yitao Cai, Yin Li, and Xiaojun Wan. 2018. Sense-aware neural models for pun location in texts. In Proceedings of ACL. ↗
  3. Jason PC Chiu and Eric Nichols. 2016. Named entity recognition with bidirectional lstm-cnns. Proceedings of TACL. ↗
  4. Samuel Doogan, Aniruddha Ghosh, Hanyang Chen, and Tony Veale. 2017. Idiom savant at semeval-2017 task 7: Detection and interpretation of english puns. In Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017).
  5. Alex Graves and Jürgen Schmidhuber. 2005. Frame-wise phoneme classification with bidirectional lstm networks. In Proceedings of 2005 IEEE International Joint Conference on Neural Networks, volume 4, pages 2047–2052. find
  6. Luheng He, Kenton Lee, Mike Lewis, and Luke Zettlemoyer. 2017. Deep semantic role labeling: What works and what’s next. In Proceedings of ACL.
  7. Xuezhe Ma and Eduard Hovy. 2016. End-to-end sequence labeling via bi-directional lstm-cnns-crf. In Proceedings of ACL. ↗
  8. Elena Mikhalkova and Yuri Karyakin. 2017. Punfields at semeval-2017 task 7: Employing roget’s thesaurus in automatic pun recognition and interpretation. In Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017). find
  9. Tristan Miller, Christian Hempelmann, and Iryna Gurevych. 2017. Semeval-2017 task 7: Detection and interpretation of english puns. In Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017).
  10. John Morkes, Hadyn K Kernal, and Clifford Nass. 1999. Effects of humor in task-oriented human-computer interaction and computer-mediated communication: A direct test of srct theory. Human-Computer Interaction, 14(4):395–435. ↗
  11. Ted Pedersen. 2017. Duluth at semeval-2017 task 7: Puns upon a midnight dreary, lexical semantics for the weak and weary. In Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017). find
  12. Christian F Hempelmann. 2008. Computational humor: Beyond the pun? The Primer of Humor Research. Humor Research, 8:333–360. ↗
  13. Hao Peng, Sam Thomson, Swabha Swayamdipta, and Noah A. Smith. 2018. Learning joint semantic parsers from disjoint data. In Proceedings of NAACL. ↗
  14. Bryan Anthony Hong and Ethel Ong. 2009. Automatically extracting word relationships as templates for pun generation. In Proceedings of the Workshop on Computational Approaches to Linguistic Creativity. find
  15. Jeffrey Pennington, Richard Socher, and Christopher Manning. 2014. Glove: Global vectors for word representation. In Proceedings of EMNLP.
  16. Vijayasaradhi Indurthi and Subba Reddy Oota. 2017. Fermi at semeval-2017 task 7: Detection and interpretation of homographic puns in english language. In Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017). ↗
  17. Aniket Pramanick and Dipankar Das. 2017. Ju cse nlp @ semeval 2017 task 7: Employing rules to detect and interpret english puns. In Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017).
  18. Arzoo Katiyar and Claire Cardie. 2016. Investigating lstms for joint extraction of opinion entities and relations. In Proceedings of ACL. find
  19. Graeme Ritchie. 2005. Computational mechanisms for pun generation. In Proceedings of the Tenth European Workshop on Natural Language Generation (ENLG-05). find
  20. John Lafferty, Andrew McCallum, and Fernando CN Pereira. 2001. Conditional random fields: Probabilistic models for segmenting and labeling sequence data. In Proceedings of ICML. find
  21. Guillaume Lample, Miguel Ballesteros, Sandeep Subramanian, Kazuya Kawakami, and Chris Dyer. 2016. Neural architectures for named entity recognition. In Proceedings of NAACL-HLT. ↗
  22. Liyuan Liu, Jingbo Shang, Xiang Ren, Frank F Xu, Huan Gui, Jian Peng, and Jiawei Han. 2018. Empower sequence labeling with task-aware neural language model. In Proceedings of AAAI.
  23. Fuli Luo, Tianyu Liu, Zexue He, Qiaolin Xia, Zhifang Sui, and Baobao Chang. 2018. Leveraging gloss knowledge in neural word sense disambiguation by hierarchical co-attention. In Proceedings of EMNLP. ↗
  24. Thorsten Schröter. 2005. Shun the Pun, Rescue the Rhyme?: The Dubbing and Subtitling of Language Play in Film. Ph.D. thesis, Estetisk-filosofiska fakulteten. ↗
  25. Özge Sevgili, Nima Ghotbi, and Selma Tekir. 2017. Nhance at semeval-2017 task 7: A computational approach using word association for puns. In Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017). find
  26. Rupesh Kumar Srivastava, Klaus Greff, and Jürgen Schmidhuber. 2015. Highway networks. arXiv preprint arXiv:1505.00387. ↗
  27. Kristina Toutanova, Dan Klein, Christopher D Manning, and Yoram Singer. 2003. Feature-rich part-of-speech tagging with a cyclic dependency network. In Proceedings of NAACL-HLT.
  28. Ankit Vadehra. 2017. Uwav at semeval-2017 task 7: Automated feature-based system for locating puns. In Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017). find
  29. Olga Vechtomova. 2017. Uwaterloo at semeval-2017 task 7: Locating the pun using syntactic characteristics and corpus-based metrics. In Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017). find
  30. Bailin Wang and Wei Lu. 2018. Learning latent opinions for aspect-level sentiment classification. In Proceedings of AAAI. find
  31. Jingjing Xu and Xu Sun. 2016. Dependency-based gated recursive neural network for chinese word segmentation. In Proceedings of ACL. find
  32. Zhiwei Yu, Jiwei Tan, and Xiaojun Wan. 2018. A neural approach to pun generation. In Proceedings of ACL. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zou-2019-joint-detection-location-english-puns/