---
title: "Hidden Sentiments — paper by Hamada (2026)"
description: "Arabic sentiment models are usually trained and tested on social-media text. 48 references."
authors: ["Hefny Abdelkader, Abdelrahman Hamada"]
year: 2026
venue: "The 7th Workshop on Open-Source Arabic Corpora and Processing Tools (OSACT7) with 5 Shared Tasks"
source: "https://aclanthology.org/2026.osact-1.1/"
doi: "10.63317/3x52cptyrpkm"
licence: "CC BY-NC 4.0 (ELRA, LREC proceedings)"
licence_url: "https://creativecommons.org/licenses/by-nc/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/abdelkader-2026-hidden-sentiments-impact-low-level/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/abdelkader-2026-hidden-sentiments-impact-low-level/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/32/) / Hidden Sentiments: The Impact of Low-level Adversarial Perturbations on Arabic Sentiment Analysis Services

# Hidden Sentiments: The Impact of Low-level Adversarial Perturbations on Arabic Sentiment Analysis Services

**Authors:** Hefny Abdelkader, Abdelrahman Hamada · **Year:** 2026 · **Venue:** [The 7th Workshop on Open-Source Arabic Corpora and Processing Tools (OSACT7) with 5 Shared Tasks](https://md.research.elamur.ai/free/en/venues/osact/) ([Workshop on Open-Source Arabic Corpora and Processing Tools with Shared Tasks 2026](https://md.research.elamur.ai/free/en/venues/osact/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY-NC 4.0 (ELRA, LREC proceedings)](https://creativecommons.org/licenses/by-nc/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.osact-1.1/) · [PDF](https://aclanthology.org/2026.osact-1.1.pdf) · [DOI 10.63317/3x52cptyrpkm](https://doi.org/10.63317/3x52cptyrpkm) · [OpenAlex](https://openalex.org/W7166506771)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
Arabic sentiment models are usually trained and tested on social-media text. That text often contains character-level noise that ordinary normalization does not remove, such as homoglyphs, stray diacritics, positional letter forms and broken or merged words. The author first scanned six widely cited Arabic SA corpora with a small Python detector to catalogue the kinds of low-level perturbation already present. Samples from three of these corpora, covering dialectal and Modern Standard Arabic, were then submitted to the black-box sentiment APIs offered by three cloud providers: Google Cloud Platform, Microsoft Azure and Amazon Web Services. Each sample was sent once in clean form and once in each of four perturbed variants. No provider beat the others in every setting, and all three proved fragile. Perturbations that left the text readable to humans cut macro-average F1 by as much as 51% relative to the clean baseline. The study matters because production systems often feed raw user content into such services. It shows that noise "hidden" in the corpora themselves can distort both research results and deployed analytics.

## Key points
- Corpora examined include OCLAR, Mawqif, AraSenti-Tweet, NADI2022 and ASTD. All were collected from social media or online platforms and carry nominal or numeric sentiment labels.
- The perturbations found in these corpora include:
  - concatenated words, split stems and detached prefixes;
  - Arabic-script and Hebrew-script homoglyphs;
  - positional (contextual) letter variants;
  - redundant Arabic diacritics and non-Arabic diacritics.
- The four attack techniques tested were homoglyph substitution, diacritic insertion, positional-variant substitution and word splitting with filler characters.
- Services were compared on macro F1, with a worst-case relative drop of 51%. GCP returns a valence score in [-1, 1], while Azure returns a confidence score in [0, 1].
- Code and analysis are public in the GitHub repository "HiddenSentiments".

## Interesting ideas
- Before building attacks, audit the existing training corpora for naturally occurring noise; real data then tells you which perturbations are realistic rather than guessed.
- Make perturbed copies of the same test set, one per perturbation type, and report each against the clean baseline so the damage can be attributed to a specific noise source.
- Keep perturbations human-readable, so that any change in prediction is clearly a model failure rather than a real change in meaning.
- Treat production systems as black boxes and probe them only through their outputs, which makes it possible to compare systems you cannot inspect.
- Split words and inserted filler marks break token boundaries even though the meaning survives, so word segmentation is itself a weak point.

## For Mova

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

- **I1** (tokens): Run a detector over our English training text for homoglyphs, zero-width characters and odd Unicode forms to decide which normalizations the tokenizer must handle.
- **I2** (learn): Add per-perturbation shadow test sets as negative controls, checking that parses and answers stay stable when only surface characters change.
- **I3** (tree): Feed gparse and the perceptron parser readability-preserving variants and flag any UD tree that changes as a robustness bug.
- **I4** (learn): Probe LLM teacher tiers and witness parsers the same way with perturbed inputs, and down-weight witnesses whose silver output flips under trivial noise.
- **I5** (tokens): Add repair rules that rejoin fragments split by stray punctuation or tatweel-like fillers before lemmatization, logging each repair so the step stays explainable.

*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

- [Carleton University](https://md.research.elamur.ai/free/en/universities/carleton-university/) — University, Ottawa, Canada · 4 works · 151 citations

📚 Shelves (2): [🖥️ 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

🧭 Topics (2): [😊 Sentiment & emotion](https://md.research.elamur.ai/free/en/topics/sentiment-emotion/) · 972 works · [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 2,908 works

## Cite

Cite as: Hefny Abdelkader, Abdelrahman Hamada. 2026. Hidden Sentiments: The Impact of Low-level Adversarial Perturbations on Arabic Sentiment Analysis Services. The 7th Workshop on Open-Source Arabic Corpora and Processing Tools (OSACT7) with 5 Shared Tasks.

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

```bibtex
@inproceedings{hamada2026hidden,
  title = {{Hidden Sentiments: The Impact of Low-level Adversarial Perturbations on Arabic Sentiment Analysis Services}},
  author = {Hefny Abdelkader, Abdelrahman Hamada},
  booktitle = {The 7th Workshop on Open-Source Arabic Corpora and Processing Tools (OSACT7) with 5 Shared Tasks},
  year = {2026},
  doi = {10.63317/3x52cptyrpkm},
  url = {https://doi.org/10.63317/3x52cptyrpkm}
}
```

- [← previous Heterogeneity in Formal Linguistic Competence of Language Models: Is Data the Real Bottleneck?](https://md.research.elamur.ai/free/en/papers/renduchintala-2026-heterogeneity-formal-linguistic-competence-language/)
- [next → HiEdit: Lifelong Model Editing with Hierarchical Reinforcement Learning](https://md.research.elamur.ai/free/en/papers/wang-2026-hiedit-lifelong-model-editing-hierarchical/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/32/)

## References

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

1. Israe Abdellaoui, Anass Ibrahimi, Mohamed Amine El Bouni, Asmaa Mourhir, Saad Driouech, and Mohamed Aghzal. 2024. Investigating offensive language detection in a low-resource setting with a robustness perspective. Big Data and Cognitive Computing, 8(12). [↗](https://doi.org/10.3390/bdcc8120170)
2. Muhammad Abdul-Mageed, Chiyu Zhang, Abdel-Rahim Elmadany, Houda Bouamor, and Nizar Habash. 2022. [NADI 2022: The third nuanced Arabic dialect identification shared task](https://md.research.elamur.ai/free/en/papers/mageed-2022-nadi-2022-third-nuanced-arabic/). In Proceedings of the Seventh Arabic Natural Language Processing Workshop (WANLP), pages 85–97, Abu Dhabi, United Arab Emirates (Hybrid). Association for Computational Linguistics.
3. Marwan Al Omari, Moustafa Al-Hajj, Nacereddine Hammami, and Amani Sabra. 2019. Sentiment classifier: Logistic regression for arabic Diacritics Neu-P RE Pos-variants Neu-P RE Split-words Neu-P RE 17% 15% 25% 0.48 0.55 0.35 25% 29% 31% 0.64 0.78 0.51 0% 0% 0% 0.56 0.68 0.48 13% 13% 6% 19% 11% 32% 0.56 0.55 0.41 11% 27% 28% 0.63 0.75 0.57 0% 0% 0% 0.53 0.56 0.36 16% 25% 37% 23% 22% 28% 0.53 0.59 0.40 27% 20% 30% 0.53 0.69 0.40 27% 7% 30% 0.54 0.55 0.40 26% 26% 30% services’ reviews in lebanon. In 2019 International Conference on Computer and Information Sciences (ICCIS), pages 1–5. [find](https://scholar.google.com/scholar?q=Marwan+Al+Omari%2C+Moustafa+Al%2DHajj%2C+Nacereddine+Hammami%2C+and+Amani+Sabra%2E+2019%2E+Sentiment+classifier%3A+Logistic+regression+for+arabic+Diacritics+Neu%2DP+RE+Pos%2Dvariants+Neu%2DP+RE+Split%2Dwords+Neu%2DP+RE+17%25+1)
4. Maged S Al-Shaibani and Irfan Ahmad. 2023. [Dotless representation of arabic text: Analysis and modeling](https://md.research.elamur.ai/free/en/papers/al-shaibani-2023-dotless-representation-arabic-text-analysis/). arXiv preprint arXiv:2312.16104. [↗](https://arxiv.org/abs/2312.16104)
5. Nora Al-Twairesh, Hend Al-Khalifa, AbdulMalik Al-Salman, and Yousef Al-Ohali. 2017. Arasenti-tweet: A corpus for arabic sentiment analysis of saudi tweets. Procedia Computer Science, 117:63–72. Arabic Computational Linguistics. [find](https://scholar.google.com/scholar?q=Nora+Al%2DTwairesh%2C+Hend+Al%2DKhalifa%2C+AbdulMalik+Al%2DSalman%2C+and+Yousef+Al%2DOhali%2E+2017%2E+Arasenti%2Dtweet%3A+A+corpus+for+arabic+sentiment+analysis+of+saudi+tweets%2E+Procedia+Computer+Science%2C+117%3A63%E2%80%9372%2E+Arabic)
6. Anwar Alajmi, Imtiaz Ahmad, and Ameer Mohammed. 2025. Evaluating the adversarial robustness of arabic spam classifiers. Neural Computing and Applications, 37(6):4323–4343. [↗](https://doi.org/10.1007/s00521-024-10778-y)
7. Nouar AlDahoul, H Abdul Karim, Mhd Adel Momo, Michael Aaron Sy, and Myles Joshua Toledo Tan. 2023. Evaluation of content moderation software for nudity and pornography detection in various scenarios. In MECON Multimedia University Engineering Conference. [find](https://scholar.google.com/scholar?q=Nouar+AlDahoul%2C+H+Abdul+Karim%2C+Mhd+Adel+Momo%2C+Michael+Aaron+Sy%2C+and+Myles+Joshua+Toledo+Tan%2E+2023%2E+Evaluation+of+content+moderation+software+for+nudity+and+pornography+detection+in+various+scenarios%2E+)
8. Hanin Alshalan and Banafsheh Rekabdar. 2023. Attacking a transformer-based models for arabic language as low resources language (lrl) using word-substitution methods. In 2023 Fifth International Conference on Transdisciplinary AI (TransAI), pages 95–101. [find](https://scholar.google.com/scholar?q=Hanin+Alshalan+and+Banafsheh+Rekabdar%2E+2023%2E+Attacking+a+transformer%2Dbased+models+for+arabic+language+as+low+resources+language+%28lrl%29+using+word%2Dsubstitution+methods%2E+In+2023+Fifth+International+Confe)
9. Basemah Alshemali. 2025a. Diacritical manipulations as adversarial attacks in arabic nlp systems. Arabian Journal for Science and Engineering. [↗](https://doi.org/10.1007/s13369-025-10266-y)
10. Basemah Alshemali. 2025b. Sentence-level adversarial examples in arabic. In Computational Science and Computational Intelligence, pages 228–242, Cham. Springer Nature Switzerland. [find](https://scholar.google.com/scholar?q=Basemah+Alshemali%2E+2025b%2E+Sentence%2Dlevel+adversarial+examples+in+arabic%2E+In+Computational+Science+and+Computational+Intelligence%2C+pages+228%E2%80%93242%2C+Cham%2E+Springer+Nature+Switzerland%2E)
11. Basemah Alshemali and Jugal Kalita. 2020. Improving the reliability of deep neural networks in nlp: A review. Knowledge-Based Systems, 191:105210. [↗](https://doi.org/10.1016/j.knosys.2019.105210)
12. Basemah Alshemali and Jugal Kalita. 2021. Character-level adversarial examples in arabic. In 2021 20th IEEE International Conference on Machine Learning and Applications (ICMLA), pages 9–14. [find](https://scholar.google.com/scholar?q=Basemah+Alshemali+and+Jugal+Kalita%2E+2021%2E+Character%2Dlevel+adversarial+examples+in+arabic%2E+In+2021+20th+IEEE+International+Conference+on+Machine+Learning+and+Applications+%28ICMLA%29%2C+pages+9%E2%80%9314%2E)
13. Nora Saleh Alturayeif, Hamzah Abdullah Luqman, and Moataz Aly Kamaleldin Ahmed. 2022. [Mawqif: A multi-label Arabic dataset for target-specific stance detection](https://md.research.elamur.ai/free/en/papers/alturayeif-2022-mawqif-multi-label-arabic-dataset/). In Proceedings of the Seventh Arabic Natural Language Processing Workshop (WANLP), pages 174–184, Abu Dhabi, United Arab Emirates (Hybrid). Association for Computational Linguistics.
14. Faisal Alvi, Mark Stevenson, and Paul Clough. 2017. Plagiarism detection in texts obfuscated with homoglyphs. In Advances in Information Retrieval, pages 669–675, Cham. Springer International Publishing. [find](https://scholar.google.com/scholar?q=Faisal+Alvi%2C+Mark+Stevenson%2C+and+Paul+Clough%2E+2017%2E+Plagiarism+detection+in+texts+obfuscated+with+homoglyphs%2E+In+Advances+in+Information+Retrieval%2C+pages+669%E2%80%93675%2C+Cham%2E+Springer+International+Publishi)
15. Sergey Berezin, Reza Farahbakhsh, and Noel Crespi. 2025. Evading toxicity detection with ASCII-art: A benchmark of spatial attacks on moderation systems. In Proceedings of the The 9th Workshop on Online Abuse and Harms (WOAH), pages 153–162, Vienna, Austria. Association for Computational Linguistics. [↗](https://doi.org/10.48550/arxiv.2409.18708)
16. Nicholas Boucher, Ilia Shumailov, Ross Anderson, and Nicolas Papernot. 2022. Bad characters: Imperceptible nlp attacks. In 2022 IEEE Symposium on Security and Privacy (SP), pages 1987– 2004. [find](https://scholar.google.com/scholar?q=Nicholas+Boucher%2C+Ilia+Shumailov%2C+Ross+Anderson%2C+and+Nicolas+Papernot%2E+2022%2E+Bad+characters%3A+Imperceptible+nlp+attacks%2E+In+2022+IEEE+Symposium+on+Security+and+Privacy+%28SP%29%2C+pages+1987%E2%80%93+2004%2E)
17. Daniel Braun, Adrian Hernandez Mendez, Florian Matthes, and Manfred Langen. 2017. Evaluating natural language understanding services for conversational question answering systems. In Proceedings of the 18th Annual SIGdial Meeting on Discourse and Dialogue, pages 174–185, Saarbrücken, Germany. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Daniel+Braun%2C+Adrian+Hernandez+Mendez%2C+Florian+Matthes%2C+and+Manfred+Langen%2E+2017%2E+Evaluating+natural+language+understanding+services+for+conversational+question+answering+systems%2E+In+Proceedings+of+th)
18. Portia Cooper, Mihai Surdeanu, and Eduardo Blanco. 2023. Hiding in plain sight: Tweets with hate speech masked by homoglyphs. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 2922–2929, Singapore. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Portia+Cooper%2C+Mihai+Surdeanu%2C+and+Eduardo+Blanco%2E+2023%2E+Hiding+in+plain+sight%3A+Tweets+with+hate+speech+masked+by+homoglyphs%2E+In+Findings+of+the+Association+for+Computational+Linguistics%3A+EMNLP+2023%2C+)
19. Joana C. Costa, Tiago Roxo, Hugo Proença, and Pedro Ricardo Morais Inácio. 2024. How deep learning sees the world: A survey on adversarial attacks & defenses. IEEE Access, 12:61113– 61136. [↗](https://doi.org/10.1109/access.2024.3395118)
20. Aldan Creo and Shushanta Pudasaini. 2025. SilverSpeak: Evading AI-generated text detec-In Proceedings of tors using homoglyphs. the 1stWorkshop on GenAI Content Detection (GenAIDetect), pages 1–46, Abu Dhabi, UAE. International Conference on Computational Linguistics. [find](https://scholar.google.com/scholar?q=Aldan+Creo+and+Shushanta+Pudasaini%2E+2025%2E+SilverSpeak%3A+Evading+AI%2Dgenerated+text+detec%2DIn+Proceedings+of+tors+using+homoglyphs%2E+the+1stWorkshop+on+GenAI+Content+Detection+%28GenAIDetect%29%2C+pages+1%E2%80%9346%2C+Ab)
21. Steffen Eger, Gözde Gül Şahin, Andreas Rücklé, Ji-Ung Lee, Claudia Schulz, Mohsen Mesgar, Krishnkant Swarnkar, Edwin Simpson, and Iryna Gurevych. 2019. [Text processing like humans do: Visually attacking and shielding NLP systems](https://md.research.elamur.ai/free/en/papers/eger-2019-text-processing-like-humans-do/). 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 1634–1647, Minneapolis, Minnesota. Association for Computational Linguistics. [↗](https://doi.org/10.48550/arxiv.1903.11508)
22. Hady ElSahar and Samhaa R. El-Beltagy. 2015. Building large arabic multi-domain resources for sentiment analysis. In Computational Linguistics and Intelligent Text Processing, pages 23– 34, Cham. Springer International Publishing. [find](https://scholar.google.com/scholar?q=Hady+ElSahar+and+Samhaa+R%2E+El%2DBeltagy%2E+2015%2E+Building+large+arabic+multi%2Ddomain+resources+for+sentiment+analysis%2E+In+Computational+Linguistics+and+Intelligent+Text+Processing%2C+pages+23%E2%80%93+34%2C+Cham%2E+Spri)
23. Tatiana Ermakova, Benjamin Fabian, Elena Golimblevskaia, and Max Henke. 2023. A comparison of commercial sentiment analysis services. SN Computer Science, 4(5):477. [↗](https://doi.org/10.1007/s42979-023-01886-y)
24. Brian Formento, Chuan Sheng Foo, Luu Anh Tuan, and See Kiong Ng. 2023. Using punctuation as an adversarial attack on deep learning-based NLP systems: An empirical study. In Findings of the Association for Computational Linguistics: EACL 2023, pages 1–34, Dubrovnik, Croatia. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Brian+Formento%2C+Chuan+Sheng+Foo%2C+Luu+Anh+Tuan%2C+and+See+Kiong+Ng%2E+2023%2E+Using+punctuation+as+an+adversarial+attack+on+deep+learning%2Dbased+NLP+systems%3A+An+empirical+study%2E+In+Findings+of+the+Association)
25. Esther Gan, Yiran Zhao, Liying Cheng, Mao Yancan, Anirudh Goyal, Kenji Kawaguchi, Min-Yen Kan, and Michael Shieh. 2024. Reasoning robustness of LLMs to adversarial typographical errors. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 10449–10459, Miami, Florida, USA. Association for Computational Linguistics. [↗](https://doi.org/10.48550/arxiv.2411.05345)
26. Shreya Goyal, Sumanth Doddapaneni, Mitesh M. Khapra, and Balaraman Ravindran. 2023. A survey of adversarial defenses and robustness in nlp. ACM Comput. Surv., 55(14s). [↗](https://doi.org/10.1145/3593042)
27. Sergiu C. Ivan, Robert Ş. Győrödi, and Cornelia A. Győrödi. 2024. Sentiment analysis using amazon web services and microsoft azure. Big Data and Cognitive Computing, 8(12). [↗](https://doi.org/10.3390/bdcc8120166)
28. Jamin Rahman Jim, Md Apon Riaz Talukder, Partha Malakar, Md Mohsin Kabir, Kamruddin Nur, and M.F. Mridha. 2024. Recent advancements and challenges of nlp-based sentiment analysis: A state-of-the-art review. Natural Language Processing Journal, 6:100059. [↗](https://doi.org/10.1016/j.nlp.2024.100059)
29. Xingkun Liu, Arash Eshghi, Pawel Swietojanski, and Verena Rieser. 2021. Benchmarking Natural Language Understanding Services for Building Conversational Agents, pages 165– 183. Springer Singapore, Singapore. [find](https://scholar.google.com/scholar?q=Xingkun+Liu%2C+Arash+Eshghi%2C+Pawel+Swietojanski%2C+and+Verena+Rieser%2E+2021%2E+Benchmarking+Natural+Language+Understanding+Services+for+Building+Conversational+Agents%2C+pages+165%E2%80%93+183%2E+Springer+Singapore%2C+Sin)
30. Yanying Mao, Qun Liu, and Yu Zhang. 2024. Sentiment analysis methods, applications, and challenges: A systematic literature review. Journal of King Saud University - Computer and Information Sciences, 36(4):102048. [↗](https://doi.org/10.1016/j.jksuci.2024.102048)
31. Yassir Matrane, Faouzia Benabbou, and Nawal Sael. 2023. A systematic literature review of arabic dialect sentiment analysis. Journal of King Saud University - Computer and Information Sciences, 35(6):101570. [↗](https://doi.org/10.1016/j.jksuci.2023.101570)
32. Walaa Medhat, Ahmed Hassan, and Hoda Korashy. 2014. Sentiment analysis algorithms and applications: A survey. Ain Shams Engineering Journal, 5(4):1093–1113. [↗](https://doi.org/10.1016/j.asej.2014.04.011)
33. Mahmoud Nabil, Mohamed Aly, and Amir Atiya. 2015. [ASTD: Arabic sentiment tweets dataset](https://md.research.elamur.ai/free/en/papers/nabil-2015-astd-arabic-sentiment-tweets-dataset/). In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 2515–2519, Lisbon, Portugal. Association for Computational Linguistics. [↗](https://doi.org/10.60692/3ee2g-zjt73)
34. Ruba Obiedat, Duha Al-Darras, Esra Alzaghoul, and Osama Harfoushi. 2021. Arabic aspect-based sentiment analysis: A systematic literature review. IEEE Access, 9:152628–152645. [↗](https://doi.org/10.1109/access.2021.3127140)
35. Oumaima Oueslati, Erik Cambria, Moez Ben HajHmida, and Habib Ounelli. 2020. [A review of sentiment analysis research in arabic language](https://md.research.elamur.ai/free/en/papers/oueslati-2020-review-sentiment-analysis-research-arabic/). Future Generation Computer Systems, 112:408– 430.
36. Łukasz Pawlik. 2025. Google cloud vs. azure: sentiment analysis accuracy for polish and english across content types. Journal of Cloud Computing, 14(1):17. [↗](https://doi.org/10.1186/s13677-025-00742-z)
37. Shilin Qiu, Qihe Liu, Shijie Zhou, and Wen Huang. 2022. Adversarial attack and defense technologies in natural language processing: A survey. Neurocomputing, 492:278–307. [↗](https://doi.org/10.1016/j.neucom.2022.04.020)
38. Azzam Radman and Rehab Duwairi. 2025. Towards a robust deep learning framework for arabic sentiment analysis. Natural Language Processing, 31(2):500–534. [↗](https://doi.org/10.1017/nlp.2024.35)
39. Kui Ren, Tianhang Zheng, Zhan Qin, and Xue Liu. 2020. Adversarial attacks and defenses in deep learning. Engineering, 6(3):346–360. [↗](https://doi.org/10.1016/j.eng.2019.12.012)
40. Sahar Sadrizadeh, AmirHossein Dabiri Aghdam, Ljiljana Dolamic, and Pascal Frossard. 2023. Targeted adversarial attacks against neural machine translation. In ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1–5. [↗](https://doi.org/10.5281/zenodo.8082869)
41. Ayed Salman, Anwar Alajmi, and Imtiaz Ahmad. 2024. [Evaluation of adversarial robustness in arabic language models](https://md.research.elamur.ai/free/en/papers/alajmi-2026-evaluation-adversarial-robustness-arabic-language/). Available at SSRN 4954707. [↗](https://doi.org/10.2139/ssrn.4954707)
42. Mayur Wankhade, Annavarapu Chandra Sekhara Rao, and Chaitanya Kulkarni. 2022. A survey on sentiment analysis methods, applications, and challenges. Artificial Intelligence Review, 55(7):5731–5780. [↗](https://doi.org/10.1007/s10462-022-10144-1)
43. Terrence E White and Manjeet Rege. 2020. Sentiment analysis on google cloud platform. Issues in Information Systems, 21(2):221–228. [find](https://scholar.google.com/scholar?q=Terrence+E+White+and+Manjeet+Rege%2E+2020%2E+Sentiment+analysis+on+google+cloud+platform%2E+Issues+in+Information+Systems%2C+21%282%29%3A221%E2%80%93228%2E)
44. Yong Xie, Dakuo Wang, Pin-Yu Chen, Jinjun Xiong, Sijia Liu, and Sanmi Koyejo. 2022. A word is worth a thousand dollars: Adversarial attack on tweets fools stock predictions. arXiv preprint arXiv:2205.01094. [↗](https://arxiv.org/abs/2205.01094)
45. Han Xu, Yao Ma, Hao-Chen Liu, Debayan Deb, Hui Liu, Ji-Liang Tang, and Anil K. Jain. 2020. Adversarial attacks and defenses in images, graphs and text: A review. International Journal of Automation and Computing, 17(2):151–178. [↗](https://doi.org/10.1007/s11633-019-1211-x)
46. Xinze Zhang, Junzhe Zhang, Zhenhua Chen, and Kun He. 2021. Crafting adversarial examples for neural machine translation. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 1967–1977, Online. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Xinze+Zhang%2C+Junzhe+Zhang%2C+Zhenhua+Chen%2C+and+Kun+He%2E+2021%2E+Crafting+adversarial+examples+for+neural+machine+translation%2E+In+Proceedings+of+the+59th+Annual+Meeting+of+the+Association+for+Computational+)
47. Jiangrui Zheng, Xueqing Liu, Mirazul Haque, Xing Qian, Guanqun Yang, and Wei Yang. 2024. HateModerate: Testing hate speech detectors against content moderation policies. In Findings of the Association for Computational Linguistics: NAACL 2024, pages 2691–2710, Mexico City, Mexico. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Jiangrui+Zheng%2C+Xueqing+Liu%2C+Mirazul+Haque%2C+Xing+Qian%2C+Guanqun+Yang%2C+and+Wei+Yang%2E+2024%2E+HateModerate%3A+Testing+hate+speech+detectors+against+content+moderation+policies%2E+In+Findings+of+the+Association)
48. Shuai Zhou, Chi Liu, Dayong Ye, Tianqing Zhu, Wanlei Zhou, and Philip S. Yu. 2022. Adversarial attacks and defenses in deep learning: From a perspective of cybersecurity. ACM Comput. Surv., 55(8). [↗](https://doi.org/10.1145/3547330)

*Updated 1 November 2026*