---
title: "The IDC System for Sentiment Classification and Sarcasm"
description: "The paper describes the IDC team's entry to the shared task on Arabic sentiment classification and… 19 references, cited by 3 papers here."
authors: ["Abraham Israeli", "Yotam Nahum", "Shai Fine", "Kfir Bar"]
year: 2021
venue: "Proceedings of the Sixth Arabic Natural Language Processing Workshop"
source: "https://aclanthology.org/2021.wanlp-1.48/"
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/israeli-2021-idc-system-sentiment-classification-sarcasm/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/israeli-2021-idc-system-sentiment-classification-sarcasm/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2021](https://md.research.elamur.ai/free/en/papers/2021/page/6/) / The IDC System for Sentiment Classification and Sarcasm Detection in Arabic

# The IDC System for Sentiment Classification and Sarcasm Detection in Arabic

**Authors:** Abraham Israeli, Yotam Nahum, Shai Fine and [Kfir Bar](https://md.research.elamur.ai/free/en/authors/kfir-bar/) · **Year:** 2021 · **Venue:** [Proceedings of the Sixth Arabic Natural Language Processing Workshop](https://md.research.elamur.ai/free/en/venues/wanlp/) ([WANLP 2021](https://md.research.elamur.ai/free/en/venues/wanlp/2021/)) · **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/2021.wanlp-1.48/) · [PDF](https://aclanthology.org/2021.wanlp-1.48.pdf) · [OpenAlex](https://openalex.org/W3153742268)

**Citations:** 7 citations (OpenAlex, 2026-10-14) · [3](https://md.research.elamur.ai/free/en/papers/israeli-2021-idc-system-sentiment-classification-sarcasm/cited-by/) in our library, of the 141,840 papers whose bibliographies we read

## Summary
The paper describes the IDC team's entry to the shared task on Arabic sentiment classification and sarcasm detection on tweets, organised by Abu Farha et al. (2021) at the Sixth Arabic NLP Workshop. Both tasks use ArSarcasm-v2, which extends ArSarcasm. Its training tweets are labelled for dialect, three-way sentiment and binary sarcasm. The test set came with dialect labels only.

The authors treat the two tasks as different problems and build a separate pipeline for each:
- **Sentiment:** a fine-tuned AraBERT with class re-weighting and FocalLoss. A gradient-boosted tree model then combines the transformer output with simple tweet meta-features and the sarcasm classifier's prediction.
- **Sarcasm:** the training data is first pruned with per-dialect topic models, then enlarged with transformer-based augmentation. Separate MARBERT models are then fine-tuned for each dialect.

On their own held-out splits the systems reach F1PN 0.75 for sentiment and F1 0.66 on the sarcastic class. These scores beat the earlier published baselines they cite. On the official test set the scores fell, and the team ranked 5th of 22 in sentiment and 9th of 27 in sarcasm. The work shows how far standard ML tricks on top of pretrained transformers carry in a low-resource, multi-dialect setting.

## Key points
- **Data:** ArSarcasm-v2 has about 12.5K training tweets, of which the 10,587 original ArSarcasm tweets are a subset, plus 3,000 test tweets. The two training subsets differ noticeably in dialect, sentiment and sarcasm distribution. MSA dominates and Maghrebi is nearly absent.
- **Sentiment pipeline:** AraBERT is trained with inverse-frequency instance weights and FocalLoss, and model selection uses F1PN. It is followed by a GBT with 30 trees of depth 3. The GBT inputs are transformer scores, text length, emoji and hashtag counts, and the sarcasm prediction. Adding the GBT lifted eval F1PN from 0.741 to 0.75. GigaBERT came out almost level with plain AraBERT.
- **Sarcasm pipeline:**
  - BERTopic-style topic models per dialect guided down-sampling of non-sarcastic tweets whose topics were poorly represented in the original ArSarcasm part. This removed 32% of the training data, 92% of it MSA.
  - MLM-based augmentation that substitutes and inserts tokens then roughly doubled the remaining corpus.
  - Four MARBERT models were trained, with Levantine and Maghrebi merged. Settings were batch 16, up to 5 epochs, and a learning rate between 4e-6 and 4e-7.
- **Errors:** the confusion matrix shows that Negative vs Neutral is the main weak spot of the sentiment model.
- **Test-set drop:** the main metric fell by 4% for sentiment and by 13.6% for sarcasm. This suggests the sarcasm gains on the internal split were partly overfit to it.

## Interesting ideas
- Each dialect model first trains for one epoch on the pooled multi-dialect data, then continues only on its own dialect's examples, so small groups borrow general signal before specialising.
- The output of one task's classifier becomes an input feature for a related task, here sarcasm feeding sentiment, through a lightweight, inspectable tree model.
- Topic models built over the data reveal topics that are over-represented relative to the target distribution, and majority-class examples from those topics are pruned instead of sampled randomly.
- The authors honestly report a large gap between their own held-out split and the hidden test set, which exposes overfitting from heavy data manipulation.
- The model checkpoint is chosen by the exact task metric (F1 over positive and negative only), not by loss or accuracy.

## For Mova

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

- **I1** (tree): gparse could pre-train on all English treebanks for one pass, then specialise to a genre or UD dialect with scarce data, such as fables or school texts.
- **I2** (frames): an irony detector in prag could emit an explicit flag that the polarity or speech-act step consumes, which keeps the reversal of literal meaning traceable.
- **I3** (learn): silver corpora could be filtered against topic or genre drift relative to the evaluation domain before training, with the removals logged so the filtering stays explainable.
- **I4** (learn): every augmentation or down-sampling step should be checked on an untouched negative-control set, so that gains are not artefacts of the reshaped data.
- **I5** (answer): reader and student models could be selected by the actual test-scoring rule of each school-course exam rather than a generic proxy.

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

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,085 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,888 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 9,296 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 9,206 works

🧭 Topics (2): [😊 Sentiment & emotion](https://md.research.elamur.ai/free/en/topics/sentiment-emotion/) · 609 works · [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 1,485 works

## Cite

Cite as: Abraham Israeli, Yotam Nahum, Shai Fine and Kfir Bar. 2021. The IDC System for Sentiment Classification and Sarcasm Detection in Arabic. Proceedings of the Sixth Arabic Natural Language Processing Workshop.

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

```bibtex
@inproceedings{israeli2021idc,
  title = {{The IDC System for Sentiment Classification and Sarcasm Detection in Arabic}},
  author = {Abraham Israeli, Yotam Nahum, Shai Fine and Kfir Bar},
  booktitle = {Proceedings of the Sixth Arabic Natural Language Processing Workshop},
  year = {2021},
  url = {https://aclanthology.org/2021.wanlp-1.48/}
}
```

- [← previous Survey on English Entity Linking on Wikidata](https://md.research.elamur.ai/free/en/papers/moller-2021-survey-english-entity-linking-wikidata/)
- [next → Adapted End-to-End Coreference Resolution System for Anaphoric Identities in Dialogues](https://md.research.elamur.ai/free/en/papers/xu-2021-adapted-end-end-coreference-resolution/)
- [↑ back to papers of 2021](https://md.research.elamur.ai/free/en/papers/2021/page/6/)

## References

19 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Muhammad Abdul-Mageed, AbdelRahim Elmadany, and El Moatez Billah Nagoudi. 2020. [Arbert & marbert: Deep bidirectional transformers for arabic](https://md.research.elamur.ai/free/en/papers/abdul-mageed-2021-arbert-marbert-deep-bidirectional-transformers/). arXiv preprint arXiv:2101.01785. [↗](https://arxiv.org/abs/2101.01785)
2. Ibrahim Abu Farha, Wajdi Zaghouani, and Walid Magdy. 2021. [Overview of the wanlp 2021 shared task on sarcasm and sentiment detection in arabic](https://md.research.elamur.ai/free/en/papers/farha-2021-overview-wanlp-2021-shared-task/). In Proceedings of the Sixth Arabic Natural Language Processing Workshop.
3. Wissam Antoun, Fady Baly, and Hazem Hajj. 2020. [Arabert: Transformer-based model for arabic language understanding](https://md.research.elamur.ai/free/en/papers/antoun-2020-arabert-transformer-based-model-arabic/). arXiv preprint arXiv:2003.00104. [↗](https://arxiv.org/abs/2003.00104)
4. Santosh Kumar Bharti, Bakhtyar Vachha, RK Pradhan, Korra Sathya Babu, and Sanjay Kumar Jena. 2016. Sarcastic sentiment detection in tweets streamed in real time: a big data approach. Digital Communications and Networks, 2(3):108–121. [↗](https://doi.org/10.1016/j.dcan.2016.06.002)
5. Houda Bouamor, Nizar Habash, Mohammad Salameh, Wajdi Zaghouani, Owen Rambow, Dana Abdulrahim, Ossama Obeid, Salam Khalifa, Fadhl Eryani, Alexander Erdmann, et al. 2018. The madar arabic dialect corpus and lexicon. In Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018). [find](https://scholar.google.com/scholar?q=Houda+Bouamor%2C+Nizar+Habash%2C+Mohammad+Salameh%2C+Wajdi+Zaghouani%2C+Owen+Rambow%2C+Dana+Abdulrahim%2C+Ossama+Obeid%2C+Salam+Khalifa%2C+Fadhl+Eryani%2C+Alexander+Erdmann%2C+et+al%2E+2018%2E+The+madar+arabic+dialect+corpus)
6. Mondher Bouazizi and Tomoaki Otsuki Ohtsuki. 2016. A pattern-based approach for sarcasm detection on twitter. IEEE Access, 4:5477–5488. [↗](https://doi.org/10.1109/access.2016.2594194)
7. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. [Bert: Pre-training of deep bidirectional transformers for language understanding](https://md.research.elamur.ai/free/en/papers/devlin-2019-bert-pre-training-deep-bidirectional/). arXiv preprint arXiv:1810.04805. [↗](https://arxiv.org/abs/1810.04805)
8. Ibrahim Abu Farha and Walid Magdy. 2019. [Mazajak: An online arabic sentiment analyser](https://md.research.elamur.ai/free/en/papers/farha-2019-mazajak-online-arabic-sentiment-analyser/). In Proceedings of the Fourth Arabic Natural Language Processing Workshop, pages 192–198.
9. Ibrahim Abu Farha and Walid Magdy. 2020. [From arabic sentiment analysis to sarcasm detection: The arsarcasm dataset](https://md.research.elamur.ai/free/en/papers/farha-2020-arabic-sentiment-analysis-sarcasm-detection/). In Proceedings of the 4th Workshop on Open-Source Arabic Corpora and Processing Tools, with a Shared Task on Offensive Language Detection, pages 32–39.
10. Jerome H Friedman. 2002. Stochastic gradient boosting. Computational statistics & data analysis, 38(4):367–378. [↗](https://doi.org/10.1016/s0167-9473(01)00065-2)
11. Maarten Grootendorst. 2020. Bertopic: Leveraging bert and c-tf-idf to create easily interpretable topics. [find](https://scholar.google.com/scholar?q=Maarten+Grootendorst%2E+2020%2E+Bertopic%3A+Leveraging+bert+and+c%2Dtf%2Didf+to+create+easily+interpretable+topics%2E)
12. Aditya Joshi, Pushpak Bhattacharyya, and Mark J Carman. 2017. [Automatic sarcasm detection: A survey](https://md.research.elamur.ai/free/en/papers/joshi-2016-automatic-sarcasm-detection-survey/). ACM Computing Surveys (CSUR), 50(5):1–22. [↗](https://doi.org/10.48550/arxiv.1602.03426)
13. Wuwei Lan, Yang Chen, Wei Xu, and Alan Ritter. 2020. Gigabert: Zero-shot transfer learning from english to arabic. In Proceedings of The 2020 Conference on Empirical Methods on Natural Language Processing (EMNLP). [find](https://scholar.google.com/scholar?q=Wuwei+Lan%2C+Yang+Chen%2C+Wei+Xu%2C+and+Alan+Ritter%2E+2020%2E+Gigabert%3A+Zero%2Dshot+transfer+learning+from+english+to+arabic%2E+In+Proceedings+of+The+2020+Conference+on+Empirical+Methods+on+Natural+Language+Proces)
14. Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollár. 2017. [Focal loss for dense object detection](https://md.research.elamur.ai/free/en/papers/lin-2017-focal-loss-dense-object-detection/). In Proceedings of the IEEE international conference on computer vision, pages 2980– 2988. [↗](https://doi.org/10.1109/iccv.2017.324)
15. Bing Liu. 2012. Sentiment analysis and opinion mining. Synthesis lectures on human language technologies, 5(1):1–167. [↗](https://doi.org/10.1007/978-3-031-02145-9)
16. Ossama Obeid, Nasser Zalmout, Salam Khalifa, Dima Taji, Mai Oudah, Bashar Alhafni, Go Inoue, Fadhl Eryani, Alexander Erdmann, and Nizar Habash. 2020. [Camel tools: An open source python toolkit for arabic natural language processing](https://md.research.elamur.ai/free/en/papers/obeid-2020-camel-tools-open-source-python/). In Proceedings of the 12th language resources and evaluation conference, pages 7022–7032.
17. Siyuan Qiu, Binxia Xu, Jie Zhang, Yafang Wang, Xiaoyu Shen, Gerard de Melo, Chong Long, and Xiaolong Li. 2020. Easyaug: An automatic textual data augmentation platform for classification tasks. In Companion Proceedings of the Web Conference 2020, pages 249–252. [find](https://scholar.google.com/scholar?q=Siyuan+Qiu%2C+Binxia+Xu%2C+Jie+Zhang%2C+Yafang+Wang%2C+Xiaoyu+Shen%2C+Gerard+de+Melo%2C+Chong+Long%2C+and+Xiaolong+Li%2E+2020%2E+Easyaug%3A+An+automatic+textual+data+augmentation+platform+for+classification+tasks%2E+In+Com)
18. Deirdre Wilson. 2006. The pragmatics of verbal irony: Echo or pretence? Lingua, 116(10):1722–1743. [↗](https://doi.org/10.1016/j.lingua.2006.05.001)
19. Lei Zhang, Shuai Wang, and Bing Liu. 2018. Deep learning for sentiment analysis: A survey. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 8(4):e1253. [↗](https://doi.org/10.1002/widm.1253)

*Updated 24 October 2026*