---
title: "DUCS at SemEval-2022 Task 6: Exploring Emojis and"
description: "The paper describes team DUCS's entry for SemEval-2022 iSarcasmEval SubTask A, which asks whether an English tweet is… 22 references."
authors: ["Vandita Grover", "Prof Hema Banati"]
year: 2022
venue: "Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022)"
source: "https://aclanthology.org/2022.semeval-1.141/"
doi: "10.18653/v1/2022.semeval-1.141"
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/grover-2022-ducs-semeval-2022-task-6/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/grover-2022-ducs-semeval-2022-task-6/)

[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/8/) / DUCS at SemEval-2022 Task 6: Exploring Emojis and Sentiments for Sarcasm Detection

# DUCS at SemEval-2022 Task 6: Exploring Emojis and Sentiments for Sarcasm Detection

**Authors:** [Vandita Grover](https://md.research.elamur.ai/free/en/authors/vandita-grover/) and Prof Hema Banati · **Year:** 2022 · **Venue:** [Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022)](https://md.research.elamur.ai/free/en/venues/semeval/) ([SemEval 2022](https://md.research.elamur.ai/free/en/venues/semeval/2022/)) · **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/2022.semeval-1.141/) · [PDF](https://aclanthology.org/2022.semeval-1.141.pdf) · [DOI 10.18653/v1/2022.semeval-1.141](https://doi.org/10.18653/v1/2022.semeval-1.141) · [OpenAlex](https://openalex.org/W4287888561)

**Citations:** 1 citations (OpenAlex, 2026-10-14)

## Summary
The paper describes team DUCS's entry for SemEval-2022 iSarcasmEval SubTask A, which asks whether an English tweet is sarcastic. Gestures and tone usually signal sarcasm, and tweets lack both. The authors therefore pulled sentiment cues from the text and from its emojis. VADER scored each tweet, keeping capitals, repeated letters and "!" as intensifiers. Each emoji was scored with the Emoji Sentiment Ranking. Emojis missing from that ranking were turned into their Unicode names and scored with VADER. The emoji scores were summed and normalized into one emoji score and polarity per tweet. Sarcastic tweets made up about a quarter of the data, so they were oversampled with nlpaug at two levels. Several models were trained with and without the sentiment features: classical ML, LSTM variants, BERT and Multimodal BERT. Multimodal BERT on the once-augmented set did best, with a test F1 of 30.71%. That placed the team 24th of 43. The study also found that sarcastic tweets tend to have positive surface sentiment and positive emojis.

## Key points
- Data: 3468 training tweets (867 sarcastic, 2601 not) with author-provided labels; the test set has 1400 tweets. Only 672 training tweets contain emojis, 195 of them sarcastic.
- Text sentiment comes from the VADER compound score. Polarity is negative at or below -0.5, positive at or above 0.5, and neutral in between. Emoji totals are normalized as score/√(score²+α) with α=15. Tweets without emojis get score 0 and neutral polarity.
- Augmentation used nlpaug with contextual BERT embeddings. Aug1 adds one synthetic copy per sarcastic tweet (about 2:1), and Aug2 adds two (1:1). Aug2 overfit, and Aug1 gave the best test score.
- Models tested:
  - Gaussian NB, Logistic Regression and SVM
  - a dense Sequential model, LSTM and Bi-LSTM with 100-d GloVe
  - BERT base uncased
  - Multimodal BERT, which also takes the tabular sentiment features
- Adding the sentiment features gave only small gains. Validation F1 was often far above test F1. Only 18 sarcastic test tweets had emojis, and F1 on emoji tweets (21.05%) was lower than on emoji-free ones (31.02%).

## Interesting ideas
- Score the words and the emojis as two separate sentiment channels, so that a mismatch between them can serve as a signal instead of being averaged away.
- For an emoji missing from the lexicon, swap it for its official Unicode name and send that text through the ordinary word-level sentiment resource, so no symbol goes unscored.
- Compute intensity features (capitals, letter repetition, exclamation marks) on the raw text before any lowercasing or normalization strips them out.
- Increasing synthetic oversampling from 2:1 to 1:1 hurt test results despite better class balance, which argues for capping augmentation and always judging it on an untouched test set.
- Report the metric separately for subsets defined by the feature under study (tweets with versus without emojis), which shows when a claimed signal rests on very few examples.

## For Mova

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

- **I1** (frames): The irony detector in prag could compare the polarity of the propositional content with that of attached emoji or interjection tokens, and record any mismatch as an explicit, explainable incongruity cue.
- **I2** (tokens): The en tokenizer could give each emoji token a lemma equal to its Unicode description, letting downstream lexicons and VerbNet/WordNet lookups treat it like a regular word.
- **I3** (tokens): Store these surface traits as token-level features (for example Emph=Caps or Elongated) next to the normalized lemma, so pragmatics can still read stress cues after normalization.
- **I4** (learn): When silver or LLM-generated examples are added, track gains on a frozen gold set to show where extra synthetic volume starts to overfit.
- **I5** (learn): Evaluations could break scores down by phenomenon buckets (e.g., sentences with or without irony markers), giving a negative-control view of whether a module actually helps where it should.

*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

- [University of Delhi](https://md.research.elamur.ai/free/en/universities/university-of-delhi/) — University, New Delhi, India · 5 works · 15 citations

📚 Shelves (4): [🖥️ 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

🧭 Topics (2): [English](https://md.research.elamur.ai/free/en/topics/english/) · 2,263 works · [😊 Sentiment & emotion](https://md.research.elamur.ai/free/en/topics/sentiment-emotion/) · 972 works

## Cite

Cite as: Vandita Grover and Prof Hema Banati. 2022. DUCS at SemEval-2022 Task 6: Exploring Emojis and Sentiments for Sarcasm Detection. Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022).

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

```bibtex
@inproceedings{grover2022ducs,
  title = {{DUCS at SemEval-2022 Task 6: Exploring Emojis and Sentiments for Sarcasm Detection}},
  author = {Vandita Grover and Prof Hema Banati},
  booktitle = {Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022)},
  year = {2022},
  doi = {10.18653/v1/2022.semeval-1.141},
  url = {https://doi.org/10.18653/v1/2022.semeval-1.141}
}
```

- [← previous Detection of Negative Campaign in Israeli Municipal Elections](https://md.research.elamur.ai/free/en/papers/litvak-2022-detection-negative-campaign-israeli-municipal/)
- [next → English across Borders: Exploring Register Variation in South Asian Newspaper Editorials](https://md.research.elamur.ai/free/en/papers/ali-2022-english-across-borders-exploring-register/)
- [↑ back to papers of 2022](https://md.research.elamur.ai/free/en/papers/2022/page/8/)

## References

22 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](https://md.research.elamur.ai/free/en/papers/farha-2022-semeval-2022-task-6-isarcasmeval/). In Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022). Association for Computational Linguistics.
2. Silvio Amir, Byron C Wallace, Hao Lyu, and Paula Carvalho Mário J Silva. 2016. [Modelling context with user embeddings for sarcasm detection in social media](https://md.research.elamur.ai/free/en/papers/amir-2016-modelling-context-user-embeddings-sarcasm/). arXiv preprint arXiv:1607.00976. [↗](https://arxiv.org/abs/1607.00976)
3. Venkateswarlu Bonta and Nandhini Kumaresh2and N Janardhan. 2019. A comprehensive study on lexicon based approaches for sentiment analysis. Asian Journal of Computer Science and Technology, 8(S2):1–6. [find](https://scholar.google.com/scholar?q=Venkateswarlu+Bonta+and+Nandhini+Kumaresh2and+N+Janardhan%2E+2019%2E+A+comprehensive+study+on+lexicon+based+approaches+for+sentiment+analysis%2E+Asian+Journal+of+Computer+Science+and+Technology%2C+8%28S2%29%3A1%E2%80%936%2E)
4. Alex Graves and Jürgen Schmidhuber. 2005. Framewise phoneme classification with bidirectional lstm and other neural network architectures. Neural networks, 18(5-6):602–610. [↗](https://doi.org/10.1016/j.neunet.2005.06.042)
5. Vandita Grover. 2021. Exploiting emojis in sentiment analysis: A survey. Journal of The Institution of Engineers (India): Series B, pages 1–14. [↗](https://doi.org/10.1007/s40031-021-00620-7)
6. Clayton Hutto and Eric Gilbert. 2014. Vader: A parsimonious rule-based model for sentiment analysis of social media text. In Proceedings of the international AAAI conference on web and social media, volume 8, pages 216–225. [find](https://scholar.google.com/scholar?q=Clayton+Hutto+and+Eric+Gilbert%2E+2014%2E+Vader%3A+A+parsimonious+rule%2Dbased+model+for+sentiment+analysis+of+social+media+text%2E+In+Proceedings+of+the+international+AAAI+conference+on+web+and+social+media%2C+v)
7. Fathonah Illia, Migunani Puspita Eugenia, and Sita Aliya Rutba. 2021. Sentiment analysis on pedulilindungi application using textblob and vader library. In Proceedings of The International Conference on Data Science and Official Statistics, volume 2021, pages 278–288. [find](https://scholar.google.com/scholar?q=Fathonah+Illia%2C+Migunani+Puspita+Eugenia%2C+and+Sita+Aliya+Rutba%2E+2021%2E+Sentiment+analysis+on+pedulilindungi+application+using+textblob+and+vader+library%2E+In+Proceedings+of+The+International+Conference+)
8. Aditya Joshi, Pushpak Bhattacharyya, and Mark James Carman. 2016. [Automatic sarcasm detection: A survey](https://md.research.elamur.ai/free/en/papers/joshi-2016-automatic-sarcasm-detection-survey/). CoRR, abs/1602.03426. [↗](https://arxiv.org/abs/1602.03426)
9. Jens Lemmens, Ben Burtenshaw, Ehsan Lotfi, Ilia Markov, and Walter Daelemans. 2020. [Sarcasm detection using an ensemble approach](https://md.research.elamur.ai/free/en/papers/lemmens-2020-sarcasm-detection-ensemble-approach/). In Proceedings of the Second Workshop on Figurative Language Processing, pages 264–269.
10. P Kralj Novak, J Smailovic, B Sluban, and I Mozetic. 2015. Emoji sentiment ranking. PLoS ONE, 10(12):e0144296. [find](https://scholar.google.com/scholar?q=P+Kralj+Novak%2C+J+Smailovic%2C+B+Sluban%2C+and+I+Mozetic%2E+2015%2E+Emoji+sentiment+ranking%2E+PLoS+ONE%2C+10%2812%29%3Ae0144296%2E)
11. Endang Wahyu Pamungkas and Viviana Patti. 2018. [# nondicevosulserio at semeval-2018 task 3: Exploiting emojis and affective content for irony detection in english tweets](https://md.research.elamur.ai/free/en/papers/pamungkas-2018-nondicevosulserio-semeval-2018-task-3/). In International Workshop on Semantic Evaluation, pages 649–654. Association for Computational Linguistics.
12. Avinash Chandra Pandey, Saksham Raj Seth, and Mahima Varshney. 2019. Sarcasm detection of amazon alexa sample set. In Advances in Signal Processing and Communication, pages 559–564. Springer. [↗](https://doi.org/10.1007/978-981-13-2553-3_54)
13. Jeffrey Pennington, Richard Socher, and Christopher D Manning. 2014. [Glove: Global vectors for word representation](https://md.research.elamur.ai/free/en/papers/pennington-2014-glove-global-vectors-word-representation/). In Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP), pages 1532–1543. [↗](https://doi.org/10.3115/v1/d14-1162)
14. Aritz Perez, Pedro Larranaga, and Inaki Inza. 2006. Supervised classification with conditional gaussian networks: Increasing the structure complexity from naive bayes. International Journal of Approximate Reasoning, 43(1):1–25. [↗](https://doi.org/10.1016/j.ijar.2006.01.002)
15. Jayashree Subramanian, Varun Sridharan, Kai Shu, and Huan Liu. 2019. Exploiting emojis for sarcasm detection. In International Conference on Social Computing, Behavioral-Cultural Modeling and Prediction and Behavior Representation in Modeling and Simulation, pages 70–80. Springer. [find](https://scholar.google.com/scholar?q=Jayashree+Subramanian%2C+Varun+Sridharan%2C+Kai+Shu%2C+and+Huan+Liu%2E+2019%2E+Exploiting+emojis+for+sarcasm+detection%2E+In+International+Conference+on+Social+Computing%2C+Behavioral%2DCultural+Modeling+and+Predicti)
16. Emilio Sulis, Delia Irazú Hernández Farías, Paolo Rosso, Viviana Patti, and Giancarlo Ruffo. 2016. Figurative messages and affect in twitter: Differences between# irony,# sarcasm and# not. Knowledge-Based Systems, 108:132–143. [↗](https://doi.org/10.1016/j.knosys.2016.05.035)
17. Karthik Sundararajan and Anandhakumar Palanisamy. 2020. Multi-rule based ensemble feature selection model for sarcasm type detection in twitter. Computational intelligence and neuroscience, 2020. [↗](https://doi.org/10.1155/2020/2860479)
18. Martin Sundermeyer, Ralf Schlüter, and Hermann Ney. 2012. Lstm neural networks for language modeling. In Thirteenth annual conference of the international speech communication association. [find](https://scholar.google.com/scholar?q=Martin+Sundermeyer%2C+Ralf+Schl%C3%BCter%2C+and+Hermann+Ney%2E+2012%2E+Lstm+neural+networks+for+language+modeling%2E+In+Thirteenth+annual+conference+of+the+international+speech+communication+association%2E)
19. Ilya Sutskever, Oriol Vinyals, and Quoc V Le. 2014. [Sequence to sequence learning with neural networks](https://md.research.elamur.ai/free/en/papers/sutskever-2014-sequence-sequence-learning-neural-networks/). Advances in neural information processing systems, 27. [↗](https://doi.org/10.48550/arxiv.1409.3215)
20. The Unicode. 2022. The Unicode Consortium. [http://www.unicode.org/emoji/charts/](http://www.unicode.org/emoji/charts/) full-emoji-list.html . [Online; accessed 15-February-2022]. [↗](http://www.unicode.org/emoji/charts/)
21. Lipo Wang. 2005. Support vector machines: theory and applications, volume 177. Springer Science & Business Media. [find](https://scholar.google.com/scholar?q=Lipo+Wang%2E+2005%2E+Support+vector+machines%3A+theory+and+applications%2C+volume+177%2E+Springer+Science+%26+Business+Media%2E)
22. Raymond E Wright. 1995. Logistic regression. [find](https://scholar.google.com/scholar?q=Raymond+E+Wright%2E+1995%2E+Logistic+regression%2E)

*Updated 1 November 2026*