---
title: "Adult Content Detection on Arabic Twitter: Analysis and"
description: "The authors describe how the two classes differ and compare SVM and Multinomial Naive Bayes on… 25 references, cited by 12 papers here."
authors: ["Hamdy Mubarak", "Sabit Hassan", "Ahmed Abdelali"]
year: 2021
venue: "WANLP 2021 (EACL workshop)"
source: "https://aclanthology.org/2021.wanlp-1.14/"
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/mubarak-2021-adult-content-detection-arabic-twitter/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/mubarak-2021-adult-content-detection-arabic-twitter/)

[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/) / Adult Content Detection on Arabic Twitter: Analysis and Experiments

# Adult Content Detection on Arabic Twitter: Analysis and Experiments

**Authors:** [Hamdy Mubarak](https://md.research.elamur.ai/free/en/authors/hamdy-mubarak/), [Sabit Hassan](https://md.research.elamur.ai/free/en/authors/sabit-hassan/) and [Ahmed Abdelali](https://md.research.elamur.ai/free/en/authors/ahmed-abdelali/) · **Year:** 2021 · **Venue:** [WANLP 2021 (EACL 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-1.14/) · [PDF](https://aclanthology.org/2021.wanlp-1.14.pdf) · [OpenAlex](https://openalex.org/W3153962697)

**Citations:** 11 citations (OpenAlex, 2026-10-14) · [12](https://md.research.elamur.ai/free/en/papers/mubarak-2021-adult-content-detection-arabic-twitter/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
The paper addresses spotting Arabic Twitter accounts that spread pornographic material or sex-related advertising. Earlier Arabic work classified accounts from many collected tweets. Here the authors use very little input: profile fields plus at most one tweet. They gathered 25M Arabic tweets from March–April 2018 and kept authors whose profiles mention an Arab country, city or nationality. A native speaker then labelled the accounts, which yielded a public corpus of 50k accounts, about 6k of them Adult, each with one tweet. The authors describe how the two classes differ and compare SVM and Multinomial Naive Bayes on character and word n-grams, FastText, Multi-BERT and AraBERT. Username plus description is enough for a macro-F1 of 94.7. Adding a single tweet raises it to 96.8 with AraBERT. This matters because suspicious accounts could be flagged early, possibly before they post at all, and users could be warned.

## Key points
- Dataset: 50k accounts with one tweet each (12% Adult, 88% Non-Adult), capped to match Twitter's sharing limit. Inter-annotator agreement on a 100-item sample gave Cohen's κ = 0.94.
- Candidates were pre-labelled with an existing vulgar-tweet detector (Mubarak and Darwish, 2019), which transferred badly: only 825 of 5,854 accounts it flagged (14%) were truly Adult.
- Adult tweets are shorter, about 9 words versus 15 for normal tweets. They typically push readers toward private contact or outside media.
- Macro-F1 results against a majority baseline of 46.7:
  - username alone: 87.1 (SVM on character n-grams)
  - username + description: 94.7 (SVM on character + word n-grams)
  - one tweet alone: 88.9
  - everything combined: 96.8 (AraBERT)
  - Screen name adds almost nothing.
- Model differences: word-only n-gram SVM trails the best system by about 2 points. Character-n-gram SVMs and the BERT models land within 0.5 of each other.

## Interesting ideas
- Filtering the raw stream with a gazetteer of place names and nationality adjectives raised the share of the rare class before any human labelling, making annotation far cheaper.
- When the best model and the gold labels disagreed on Non-Adult accounts, about 70% of the disagreements turned out to be labelling mistakes. Classifier disagreement therefore works as an annotation-audit tool.
- An ablation over each information source (username, screen name, description, tweet) showed which fields carry signal and which, like screen name, are nearly empty.
- Character n-grams beat word n-grams because spammers deliberately misspell words to dodge keyword lists. Subword evidence holds up under creative spelling.
- A pre-labeller from a neighbouring task (vulgar talk) did poorly on the target task (adult ads), showing that a related task is not a reliable teacher.

## For Mova

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

- **I1** (learn): Before silver annotation, prefilter text with lexical triggers for rare constructions so that witness parsers and LLM teachers spend their effort on the cases that matter.
- **I2** (learn): Route parser-versus-treebank disagreements to an audit queue, and fix gold errors before treating them as model failures.
- **I3** (answer): Ablate the inputs a reader uses (tree, frames, world facts) to show which layer actually supports each answer, which keeps the pipeline explainable.
- **I4** (tokens): Add character-n-gram fallback features in the tagger and lemmatizer for unseen or misspelled forms, instead of mapping them to a single unknown token.
- **I5** (learn): Measure the precision of any adjacent-task teacher on a small gold sample before letting its labels shape silver data.

*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 (3): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 19,503 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 18,805 works · [🗄️ Information Systems](https://md.research.elamur.ai/free/en/shelves/computer-science/information-systems/) · 1,888 works

🧭 Topics (1): [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 3,068 works

## Cite

Cite as: Hamdy Mubarak, Sabit Hassan and Ahmed Abdelali. 2021. Adult Content Detection on Arabic Twitter: Analysis and Experiments. WANLP 2021 (EACL workshop).

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

```bibtex
@inproceedings{mubarak2021adult,
  title = {{Adult Content Detection on Arabic Twitter: Analysis and Experiments}},
  author = {Hamdy Mubarak, Sabit Hassan and Ahmed Abdelali},
  booktitle = {WANLP 2021 (EACL workshop)},
  year = {2021},
  url = {https://aclanthology.org/2021.wanlp-1.14/}
}
```

- [← previous TacticZero: Learning to Prove Theorems from Scratch with Deep Reinforcement Learning](https://md.research.elamur.ai/free/en/papers/wu-2021-tacticzero-learning-prove-theorems-scratch/)
- [next → Conundrums in Event Coreference Resolution: Making Sense of the State of the Art](https://md.research.elamur.ai/free/en/papers/lu-2021-conundrums-event-coreference-resolution-making/)
- [↑ back to papers of 2021](https://md.research.elamur.ai/free/en/papers/2021/page/6/)

## References

25 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Ahmed Abdelali, Hamdy Mubarak, Younes Samih, Sabit Hassan, and Kareem Darwish. 2020. [Arabic dialect identification in the wild](https://md.research.elamur.ai/free/en/papers/abdelali-2020-arabic-dialect-identification-wild/). arXiv preprint arXiv:2005.06557. [↗](https://arxiv.org/abs/2005.06557)
2. Ehab Abozinadah, Alex Mbaziira, and James Jr. 2015. Detection of abusive accounts with arabic tweets. volume 1. [find](https://scholar.google.com/scholar?q=Ehab+Abozinadah%2C+Alex+Mbaziira%2C+and+James+Jr%2E+2015%2E+Detection+of+abusive+accounts+with+arabic+tweets%2E+volume+1%2E)
3. Ehab A. Abozinadah and James H. Jones. 2017. A statistical learning approach to detect abusive twitter accounts. In Proceedings of the International Conference on Compute and Data Analysis, ICCDA ’17, page 6–13, New York, NY, USA. Association for Computing Machinery. [find](https://scholar.google.com/scholar?q=Ehab+A%2E+Abozinadah+and+James+H%2E+Jones%2E+2017%2E+A+statistical+learning+approach+to+detect+abusive+twitter+accounts%2E+In+Proceedings+of+the+International+Conference+on+Compute+and+Data+Analysis%2C+ICCDA+%E2%80%9917%2C)
4. 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, Florence, Italy. Association for Computational Linguistics.
5. N. Albadi, M. Kurdi, and S. Mishra. 2018. Are they our brothers? analysis and detection of religious hate speech in the arabic twittersphere. In 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM), pages 69–76. [find](https://scholar.google.com/scholar?q=N%2E+Albadi%2C+M%2E+Kurdi%2C+and+S%2E+Mishra%2E+2018%2E+Are+they+our+brothers%3F+analysis+and+detection+of+religious+hate+speech+in+the+arabic+twittersphere%2E+In+2018+IEEE%2FACM+International+Conference+on+Advances+in+S)
6. Ali Alshehri, El Moatez Billah Nagoudi, Hassan Alhuzali, and Muhammad Abdul-Mageed. 2018. Think before your click: Data and models for adult content in arabic twitter. [find](https://scholar.google.com/scholar?q=Ali+Alshehri%2C+El+Moatez+Billah+Nagoudi%2C+Hassan+Alhuzali%2C+and+Muhammad+Abdul%2DMageed%2E+2018%2E+Think+before+your+click%3A+Data+and+models+for+adult+content+in+arabic+twitter%2E)
7. Wissam Antoun, Fady Baly, and Hazem M. 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, abs/2003.00104. [↗](https://arxiv.org/abs/2003.00104)
8. H. Cheng, X. Xing, X. Liu, and Q. Lv. 2015. Isc: An iterative social based classifier for adult account detection on twitter. IEEE Transactions on Knowledge and Data Engineering, 27(4):1045–1056. [find](https://scholar.google.com/scholar?q=H%2E+Cheng%2C+X%2E+Xing%2C+X%2E+Liu%2C+and+Q%2E+Lv%2E+2015%2E+Isc%3A+An+iterative+social+based+classifier+for+adult+account+detection+on+twitter%2E+IEEE+Transactions+on+Knowledge+and+Data+Engineering%2C+27%284%29%3A1045%E2%80%931056%2E)
9. Hamdy Mubarak and Kareem Darwish. 2019. Arabic offensive language classification on twitter. In International Conference on Social Informatics, pages 269–276. Springer. [find](https://scholar.google.com/scholar?q=Hamdy+Mubarak+and+Kareem+Darwish%2E+2019%2E+Arabic+offensive+language+classification+on+twitter%2E+In+International+Conference+on+Social+Informatics%2C+pages+269%E2%80%93276%2E+Springer%2E)
10. Michael Conover, Jacob Ratkiewicz, Matthew R Francisco, Bruno Gonçalves, Filippo Menczer, and Alessandro Flammini. 2011. Political polarization on twitter. ICWSM, 133:89–96. [find](https://scholar.google.com/scholar?q=Michael+Conover%2C+Jacob+Ratkiewicz%2C+Matthew+R+Francisco%2C+Bruno+Gonc%CC%A7alves%2C+Filippo+Menczer%2C+and+Alessandro+Flammini%2E+2011%2E+Political+polarization+on+twitter%2E+ICWSM%2C+133%3A89%E2%80%9396%2E)
11. Hamdy Mubarak, Kareem Darwish, and Walid Magdy. 2017. [Abusive language detection on Arabic social media](https://md.research.elamur.ai/free/en/papers/mubarak-2017-abusive-language-detection-arabic-social/). In Proceedings of the First Workshop on Abusive Language Online, pages 52–56, Vancouver, BC, Canada. Association for Computational Linguistics.
12. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. [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/). In 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), pages 4171–4186, Minneapolis, Minnesota. Association for Computational Linguistics. [↗](https://doi.org/10.18653/v1/n19-1423)
13. Chris Grier, Kurt Thomas, Vern Paxson, and Michael Zhang. 2010. @spam: The underground on 140 characters or less. In Proceedings of the 17th ACM Conference on Computer and Communications Security, CCS ’10, page 27–37, New York, NY, USA. Association for Computing Machinery. [find](https://scholar.google.com/scholar?q=Chris+Grier%2C+Kurt+Thomas%2C+Vern+Paxson%2C+and+Michael+Zhang%2E+2010%2E+%40spam%3A+The+underground+on+140+characters+or+less%2E+In+Proceedings+of+the+17th+ACM+Conference+on+Computer+and+Communications+Security%2C+CCS)
14. Sabit Hassan, Younes Samih, Hamdy Mubarak, and Ahmed Abdelali. 2020a. [ALT at SemEval-2020 task 12: Arabic and English offensive language identification in social media](https://md.research.elamur.ai/free/en/papers/hassan-2020-alt-semeval-2020-task-12/). In Proceedings of the Fourteenth Workshop on Semantic Evaluation, pages 1891–1897, Barcelona (online). International Committee for Computational Linguistics.
15. Sabit Hassan, Younes Samih, Hamdy Mubarak, Ahmed Abdelali, Ammar Rashed, and Shammur Chowdhury. 2020b. [Alt submission for osact shared task on offensive language detection](https://md.research.elamur.ai/free/en/papers/hassan-2020-alt-submission-osact-shared-task/). In Proceedings of the 4th Workshop on Open-Source Arabic Corpora and Processing Tools (LREC 2020), page 61–65.
16. Wafa Herzallah, Hossam Faris, and Omar Adwan. 2018. Feature engineering for detecting spammers on twitter: Modelling and analysis. Journal of Information Science, 44(2):230–247. [find](https://scholar.google.com/scholar?q=Wafa+Herzallah%2C+Hossam+Faris%2C+and+Omar+Adwan%2E+2018%2E+Feature+engineering+for+detecting+spammers+on+twitter%3A+Modelling+and+analysis%2E+Journal+of+Information+Science%2C+44%282%29%3A230%E2%80%93247%2E)
17. Armand Joulin, Edouard Grave, Piotr Bojanowski, Matthijs Douze, Hervé Jégou, and Tomas Mikolov. 2016. [Fasttext.zip: Compressing text classification models](https://md.research.elamur.ai/free/en/papers/joulin-2016-fasttext-zip-compressing-text-classification/). CoRR, abs/1612.03651. [↗](https://arxiv.org/abs/1612.03651)
18. J. Richard Landis and Gary G. Koch. 1977. The measurement of observer agreement for categorical data. Biometrics, 33(1):159–174. [find](https://scholar.google.com/scholar?q=J%2E+Richard+Landis+and+Gary+G%2E+Koch%2E+1977%2E+The+measurement+of+observer+agreement+for+categorical+data%2E+Biometrics%2C+33%281%29%3A159%E2%80%93174%2E)
19. Arun S. Maiya. 2020. ktrain: A low-code library for augmented machine learning. arXiv, arXiv:2004.10703 [cs.LG]. [↗](https://arxiv.org/abs/2004.10703)
20. Kimberly J Mitchell, David Finkelhor, and Janis Wolak. 2003. The exposure of youth to unwanted sexual material on the internet: A national survey of risk, impact, and prevention. Youth & Society, 34(3):330– 358. [find](https://scholar.google.com/scholar?q=Kimberly+J+Mitchell%2C+David+Finkelhor%2C+and+Janis+Wolak%2E+2003%2E+The+exposure+of+youth+to+unwanted+sexual+material+on+the+internet%3A+A+national+survey+of+risk%2C+impact%2C+and+prevention%2E+Youth+%26+Society%2C+34%283)
21. Hamdy Mubarak, Ahmed Abdelali, Sabit Hassan, and Kareem Darwish. 2020. Spam detection on arabic twitter. In Social Informatics, pages 237–251, Cham. Springer International Publishing. [find](https://scholar.google.com/scholar?q=Hamdy+Mubarak%2C+Ahmed+Abdelali%2C+Sabit+Hassan%2C+and+Kareem+Darwish%2E+2020%2E+Spam+detection+on+arabic+twitter%2E+In+Social+Informatics%2C+pages+237%E2%80%93251%2C+Cham%2E+Springer+International+Publishing%2E)
22. Po-Ching Lin and Po-Min Huang. 2013. A study of effective features for detecting long-surviving twitter spam accounts. In 2013 15th International Conference on Advanced Communications Technology (ICACT), pages 841–846. [find](https://scholar.google.com/scholar?q=Po%2DChing+Lin+and+Po%2DMin+Huang%2E+2013%2E+A+study+of+effective+features+for+detecting+long%2Dsurviving+twitter+spam+accounts%2E+In+2013+15th+International+Conference+on+Advanced+Communications+Technology+%28ICAC)
23. Monika Singh, Divya Bansal, and Sanjeev Sofat. 2016. Behavioral analysis and classification of spammers distributing pornographic content in social media. Social Network Analysis and Mining, 6. [find](https://scholar.google.com/scholar?q=Monika+Singh%2C+Divya+Bansal%2C+and+Sanjeev+Sofat%2E+2016%2E+Behavioral+analysis+and+classification+of+spammers+distributing+pornographic+content+in+social+media%2E+Social+Network+Analysis+and+Mining%2C+6%2E)
24. Chao Yang, Robert Harkreader, and Guofei Gu. 2013. Empirical evaluation and new design for fighting evolving twitter spammers. IEEE Transactions on Information Forensics and Security, 8(8):1280– 1293. [find](https://scholar.google.com/scholar?q=Chao+Yang%2C+Robert+Harkreader%2C+and+Guofei+Gu%2E+2013%2E+Empirical+evaluation+and+new+design+for+fighting+evolving+twitter+spammers%2E+IEEE+Transactions+on+Information+Forensics+and+Security%2C+8%288%29%3A1280%E2%80%93+1293%2E)
25. Harish Yenala, Ashish Jhanwar, Manoj Chinnakotla, and Jay Goyal. 2017. Deep learning for detecting inappropriate content in text. International Journal of Data Science and Analytics. [find](https://scholar.google.com/scholar?q=Harish+Yenala%2C+Ashish+Jhanwar%2C+Manoj+Chinnakotla%2C+and+Jay+Goyal%2E+2017%2E+Deep+learning+for+detecting+inappropriate+content+in+text%2E+International+Journal+of+Data+Science+and+Analytics%2E)

*Updated 4 November 2026*