---
title: "YNU-HPCC at SemEval-2022 Task 6 — Zheng 2022"
description: "The paper describes the YNU-HPCC entry to SemEval-2022 Task 6 (iSarcasmEval)."
authors: ["Guangmin Zheng", "Jin Wang", "Xuejie Zhang"]
year: 2022
venue: "Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022)"
source: "https://aclanthology.org/2022.semeval-1.134/"
doi: "10.18653/v1/2022.semeval-1.134"
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/zheng-2022-ynu-hpcc-semeval-2022-task/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/zheng-2022-ynu-hpcc-semeval-2022-task/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2022](https://md.research.elamur.ai/free/en/papers/2022/page/4/) / YNU-HPCC at SemEval-2022 Task 6: Transformer-based Model for Intended Sarcasm Detection in English and Arabic

# YNU-HPCC at SemEval-2022 Task 6: Transformer-based Model for Intended Sarcasm Detection in English and Arabic

**Authors:** Guangmin Zheng, Jin Wang, Xuejie Zhang · **Institutions:** Yunnan University · **Year:** 2022 · **Venue:** Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2022.semeval-1.134/) · [PDF](https://aclanthology.org/2022.semeval-1.134.pdf) · [DOI 10.18653/v1/2022.semeval-1.134](https://doi.org/10.18653/v1/2022.semeval-1.134) · [OpenAlex](https://openalex.org/W4287854975)

**Citations:** 4 citations (OpenAlex, 2026-10-14) · [5](https://md.research.elamur.ai/free/en/papers/zheng-2022-ynu-hpcc-semeval-2022-task/cited-by/) in our library, of the 140884 papers whose bibliographies we read

## Summary
The paper describes the YNU-HPCC entry to SemEval-2022 Task 6 (iSarcasmEval). The task asks three things about tweets in English and Arabic: whether a tweet is sarcastic (A), which of six ironic speech types it shows (B, English only), and which member of a pair is the sarcastic original rather than its literal rewording (C). The team used only the organizers' data: 3467 English and 2602 Arabic items. Tweets were cleaned with regular expressions, with user handles, URLs and escape characters removed and hashtags kept intact. They were then encoded with bert-base-multilingual-uncased or roberta-base, and a small classifier was placed on top of the [CLS] vector. For B, a single network predicted all labels at once. Its loss was binary cross-entropy weighted toward rare labels, and each label was accepted when its sigmoid score was at least 0.5. The best results ranked 12/43 in English A, 11/22 in B and 3/16 in English C. The work shows how far plain fine-tuning goes on sarcasm and how weak it remains on fine-grained irony types.

## Key points
- Official rankings were English (Arabic): A 12/43 (21/32), B 11/22, C 3/16 (8/13).
- In English, roberta-base gave the best scores: F1 0.392 on A and accuracy 0.860 on C.
- For Arabic, fine-tuning multilingual BERT on English plus Arabic data raised scores over Arabic-only training: A F1 rose from 0.202 to 0.245, and C accuracy rose from 0.5 to 0.755.
- Subtask B labels were heavily skewed: sarcasm 713, irony 155, rhetorical_question 101, overstatement 40, satire 25, understatement 10.
- In B, BCE with per-label weights reached 0.0795 macro-F1. This beat plain BCE (0.0672), six separate binary classifiers (0.0702), Focal (0.0646) and Dice (0.0379), although all scores stay very low.
- Training used AdamW, learning rate 2e-5, batch size 16, dropout 0.1 and an 8:2 train/dev split. Classes were sampled so that the two binary labels appeared in similar numbers.

## Interesting ideas
- For the pair task, every tweet/rewording pair is used twice, once in each order with opposite labels. This doubles the data and stops the model from learning that position predicts the answer. — For Mova (learn): any pairwise judge trained on silver data, such as "which paraphrase keeps the meaning" or "which parse is better", could use order-swapped copies to remove position bias.
- A literal restatement serves as a contrast partner for an ironic sentence. Sarcasm is detected by comparing what was said against a plain version, not from the sentence alone. — For Mova (frames): the irony module in prag could build the literal frame reading and compare its polarity with context, marking irony where the two clash. This keeps the decision explainable.
- Loss weights for each label are derived from how often the label occurs. This lets a single multi-output head attend to tiny categories like understatement without training separate models. — For Mova (learn): rare speech-act or semantic-role labels in teacher-tier training sets could get similar frequency-based weights so that frequent labels do not swamp them.
- Hashtags are kept as whole units during cleaning, on the view that they carry much of the tweet's intent. Handles and URLs, which carry little, are removed. — For Mova (tokens): the en tokenizer could treat hashtags and similar markup as single tokens with a dedicated tag rather than splitting them into punctuation and words.
- Several loss functions and set-ups are compared side by side on the same split, and the six-binary baseline acts as a reference point. Together these show that tiny macro-F1 gains need context. — For Mova (learn): reporting each new training trick next to simple baselines, as negative controls, helps tell real gains from noise when scores are near the floor.

*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: [Linguistics & languages](https://md.research.elamur.ai/free/en/shelves/linguistics/)

Topics: [Sentiment & emotion](https://md.research.elamur.ai/free/en/topics/sentiment-emotion/) · [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/)

## Cite

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

```bibtex
@inproceedings{zheng2022ynu,
  title = {{YNU-HPCC at SemEval-2022 Task 6: Transformer-based Model for Intended Sarcasm Detection in English and Arabic}},
  author = {Zheng, Guangmin and Wang, Jin and Zhang, Xuejie},
  booktitle = {Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022)},
  year = {2022},
  doi = {10.18653/v1/2022.semeval-1.134},
  url = {https://doi.org/10.18653/v1/2022.semeval-1.134}
}
```

- [← previous The Index Thomisticus Treebank as Linked Data in the LiLa Knowledge Base](https://md.research.elamur.ai/free/en/papers/mambrini-2022-index-thomisticus-treebank-as-linked/)
- [next → Automatic Dialect Density Estimation for African American English](https://md.research.elamur.ai/free/en/papers/johnson-2022-automatic-dialect-density-estimation-african/)
- [↑ back to papers of 2022](https://md.research.elamur.ai/free/en/papers/2022/page/4/)

## References

15 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Ibrahim Abu Farha, Silviu Oprea, Steven Wilson, and Walid Magdy. 2022. SemEval-2022 Task 6: iSarcasmEval, Intended Sarcasm Detection in English and Arabic. In Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022). Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Ibrahim+Abu+Farha%2C+Silviu+Oprea%2C+Steven+Wilson%2C+and+Walid+Magdy%2E+2022%2E+SemEval%2D2022+Task+6%3A+iSarcasmEval%2C+Intended+Sarcasm+Detection+in+English+and+Arabic%2E+In+Proceedings+of+the+16th+International+Wor) [our card](https://md.research.elamur.ai/free/en/papers/farha-2022-semeval-2022-task-6-isarcasmeval/)
2. David Bamman and Noah A. Smith. 2015. Contextualized sarcasm detection on twitter. In Proceedings of the 9th International Conference on Web and Social Media (ICWSM 2015), pages 574–577. [find](https://scholar.google.com/scholar?q=David+Bamman+and+Noah+A%2E+Smith%2E+2015%2E+Contextualized+sarcasm+detection+on+twitter%2E+In+Proceedings+of+the+9th+International+Conference+on+Web+and+Social+Media+%28ICWSM+2015%29%2C+pages+574%E2%80%93577%2E)
3. Jiaxu Dao, Jin Wang, and Xuejie Zhang. 2020. YNU-HPCC at SemEval-2020 task 11: LSTM network for detection of propaganda techniques in news articles. In Proceedings of the Fourteenth Workshop on Semantic Evaluation, pages 1509–1515, Barcelona (online). International Committee for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Jiaxu+Dao%2C+Jin+Wang%2C+and+Xuejie+Zhang%2E+2020%2E+YNU%2DHPCC+at+SemEval%2D2020+task+11%3A+LSTM+network+for+detection+of+propaganda+techniques+in+news+articles%2E+In+Proceedings+of+the+Fourteenth+Workshop+on+Semant)
4. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805. [↗](https://arxiv.org/abs/1810.04805) [our card](https://md.research.elamur.ai/free/en/papers/devlin-2019-bert-pre-training-deep-bidirectional/)
5. Aniruddha Ghosh and Tony Veale. 2016. Fracking sarcasm using neural network. In Proceedings of the 7th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis, pages 161–169, San Diego, California. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Aniruddha+Ghosh+and+Tony+Veale%2E+2016%2E+Fracking+sarcasm+using+neural+network%2E+In+Proceedings+of+the+7th+Workshop+on+Computational+Approaches+to+Subjectivity%2C+Sentiment+and+Social+Media+Analysis%2C+pages+) [our card](https://md.research.elamur.ai/free/en/papers/ghosh-2016-fracking-sarcasm-neural-network/)
6. Roberto González-Ibáñez, Smaranda Muresan, and Nina Wacholder. 2011. Identifying sarcasm in Twitter: A closer look. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, pages 581–586, Portland, Oregon, USA. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Roberto+Gonz%C3%A1lez%2DIb%C3%A1%C3%B1ez%2C+Smaranda+Muresan%2C+and+Nina+Wacholder%2E+2011%2E+Identifying+sarcasm+in+Twitter%3A+A+closer+look%2E+In+Proceedings+of+the+49th+Annual+Meeting+of+the+Association+for+Computational+Lingu) [our card](https://md.research.elamur.ai/free/en/papers/ibanez-2011-identifying-sarcasm-twitter-closer-look/)
7. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv: 1907.11692. [↗](https://arxiv.org/abs/1907.11692)
8. Ilya Loshchilov and Frank Hutter. 2019. Decoupled weight decay regularization. arXiv preprint arXiv: 1711.05101. [↗](https://arxiv.org/abs/1711.05101)
9. Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Efficient estimation of word representations in vector space. arXiv preprint arXiv: 1301.3781. [↗](https://arxiv.org/abs/1301.3781) [our card](https://md.research.elamur.ai/free/en/papers/mikolov-2013-efficient-estimation-word-representations-vector/)
10. Shabnam Nazmi, Xuyang Yan, Abdollah Homaifar, and Emily Doucette. 2020. Evolving multi-label classification rules by exploiting high-order label correlations. Neurocomputing, 417:176–186. [find](https://scholar.google.com/scholar?q=Shabnam+Nazmi%2C+Xuyang+Yan%2C+Abdollah+Homaifar%2C+and+Emily+Doucette%2E+2020%2E+Evolving+multi%2Dlabel+classification+rules+by+exploiting+high%2Dorder+label+correlations%2E+Neurocomputing%2C+417%3A176%E2%80%93186%2E)
11. Bo Peng, Jin Wang, and Xuejie Zhang. 2018. YNU-HPCC at SemEval-2018 task 3: Ensemble neural network models for irony detection on Twitter. In Proceedings of the 12th International Workshop on Semantic Evaluation, pages 622–627, New Orleans, Louisiana. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Bo+Peng%2C+Jin+Wang%2C+and+Xuejie+Zhang%2E+2018%2E+YNU%2DHPCC+at+SemEval%2D2018+task+3%3A+Ensemble+neural+network+models+for+irony+detection+on+Twitter%2E+In+Proceedings+of+the+12th+International+Workshop+on+Semantic)
12. Telmo Pires, Eva Schlinger, and Dan Garrette. 2019. How multilingual is multilingual BERT? In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4996–5001, Florence, Italy. Association for Computational Linguistics. [↗](https://doi.org/10.18653/v1/p19-1493) [our card](https://md.research.elamur.ai/free/en/papers/pires-2019-how-multilingual-is-multilingual-bert/)
13. Grigorios Tsoumakas, Ioannis Katakis, and Ioannis Vla-havas. 2011. Random k-labelsets for multilabel classification. IEEE Transactions on Knowledge and Data Engineering, 23(7):1079–1089. [find](https://scholar.google.com/scholar?q=Grigorios+Tsoumakas%2C+Ioannis+Katakis%2C+and+Ioannis+Vla%2Dhavas%2E+2011%2E+Random+k%2Dlabelsets+for+multilabel+classification%2E+IEEE+Transactions+on+Knowledge+and+Data+Engineering%2C+23%287%29%3A1079%E2%80%931089%2E)
14. Meishan Zhang, Yue Zhang, and Guohong Fu. 2016. Tweet sarcasm detection using deep neural network. In Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 2449–2460, Osaka, Japan. [find](https://scholar.google.com/scholar?q=Meishan+Zhang%2C+Yue+Zhang%2C+and+Guohong+Fu%2E+2016%2E+Tweet+sarcasm+detection+using+deep+neural+network%2E+In+Proceedings+of+COLING+2016%2C+the+26th+International+Conference+on+Computational+Linguistics%3A+Techni) [our card](https://md.research.elamur.ai/free/en/papers/zhang-2016-tweet-sarcasm-detection-deep-neural/)
15. The COLING 2016 Organizing Committee. [find](https://scholar.google.com/scholar?q=The+COLING+2016+Organizing+Committee%2E)