---
title: "ASTD: Arabic Sentiment Tweets Dataset — Nabil 2015"
description: "The paper addresses the shortage of standard, publicly available benchmarks for Arabic sentiment analysis on social media."
authors: ["Mahmoud Nabil", "Mohamed Aly", "Amir Atiya"]
year: 2015
venue: "Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing"
source: "https://aclanthology.org/D15-1299/"
doi: "10.18653/v1/d15-1299"
licence: "CC BY-NC-SA 3.0 (ACL Anthology, before 2016)"
licence_url: "https://creativecommons.org/licenses/by-nc-sa/3.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/nabil-2015-astd-arabic-sentiment-tweets-dataset/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/nabil-2015-astd-arabic-sentiment-tweets-dataset/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2015](https://md.research.elamur.ai/free/en/papers/2015/) / ASTD: Arabic Sentiment Tweets Dataset

# ASTD: Arabic Sentiment Tweets Dataset

**Authors:** Mahmoud Nabil, Mohamed Aly, Amir Atiya · **Year:** 2015 · **Venue:** Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing · **Licence:** [CC BY-NC-SA 3.0 (ACL Anthology, before 2016)](https://creativecommons.org/licenses/by-nc-sa/3.0/)

**Links:** [Publisher's page](https://aclanthology.org/D15-1299/) · [PDF](https://aclanthology.org/D15-1299.pdf) · [DOI 10.18653/v1/d15-1299](https://doi.org/10.18653/v1/d15-1299) · [OpenAlex](https://openalex.org/W2250594687)

**Citations:** 387 citations (openalex-snapshot · 2026-10-14) · [83](#cited-by) in our library, of the 140884 papers whose bibliographies we read

## Summary
The paper addresses the shortage of standard, publicly available benchmarks for Arabic sentiment analysis on social media. The authors build ASTD, a corpus of roughly 10k Egyptian-oriented Arabic tweets. The raw tweets come from two sources: the timelines of highly active Egyptian accounts and tweets retrieved through trending Egyptian hashtags. Workers on Amazon Mechanical Turk labeled each tweet into one of four categories: objective, subjective positive, subjective negative, or subjective mixed. Only tweets on which raters agreed were kept. The authors describe the corpus statistics and release fixed train/validation/test splits in both a balanced and an unbalanced version. They then benchmark a range of standard scikit-learn classifiers on n-gram count and Tf-Idf features. Four-way classification proves hard. On the balanced split the best models reach only about 0.49 weighted accuracy, while the unbalanced split yields roughly 0.69 accuracy and about 0.63 weighted F1. The released data and code give later Arabic social-media sentiment work a common baseline to compare against.

## Key points
- **Collection:** About 84,000 tweets were collected in two ways. Roughly 36k came from 30 top Egyptian accounts identified via SocialBakers, and about 48k came from around 2,500 hashtags crawled from EgyptTrends. Filtering and cleanup left 54,716 Arabic tweets.
- **Annotation:** Labels were collected on AMT through the Boto API. A tweet was accepted when at least two of three raters agreed, which produced 10,006 conflict-free tweets.
- **Class distribution:** 6,691 objective, 1,684 negative, 832 mixed, and 799 positive tweets.
- **Corpus size:** 160,206 tokens and 38,743 distinct vocabulary items. Tweets average and median about 16 tokens, with a maximum of 45.
- **Splits and evaluation:**
  - The data is split 6:2:2 into train, validation, and test.
  - The balanced version uses 481 training tweets per class and 159 per class for validation and for test.
  - Eight classifiers were compared: multinomial (MNB) and Bernoulli (BNB) naive Bayes, a support vector machine (SVM), logistic regression, stochastic gradient descent (SGD), a passive-aggressive model, a perceptron, and k-nearest neighbours (KNN).
  - Features: 1g up to 1g+2g+3g, with and without Tf-Idf.
  - Models are evaluated on the test set after training on train plus validation.
  - The authors name SVM as the most reliable choice. In the reported table, several linear models perform similarly.

## Interesting ideas
- **Two-pronged seeding of tweet collection.** The authors combined influential accounts with trending hashtags. This covers both steady high-volume voices and topical bursts, and the approach could be reused to build a regional social-media corpus for any language or country.
- **Majority-agreement filtering of crowd labels.** Keeping only tweets where at least two of three annotators agree is a cheap way to obtain cleaner labels from noisy crowdworkers. The cost is losing some hard examples, so it suits projects that prioritize label reliability over coverage.
- **Separate mixed-sentiment class.** Mixed opinions are labeled as their own category rather than forced into positive or negative. This makes the dataset more realistic for social text, where people often express ambivalence.
- **Balanced and unbalanced versions of the same splits.** Releasing both lets researchers separate the effect of class skew from the effect of training-set size. The authors attribute the weaker balanced results mainly to having less data, which shows why both views are informative.
- **Broad baseline grid on a new dataset.** The authors ran many simple classifiers across n-gram orders and weighting schemes, with or without Tf-Idf. This gives future work a transparent, reproducible baseline and shows how sensitive results are to feature choices, not just a single headline number.

*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: [Natural language processing](https://md.research.elamur.ai/free/en/shelves/nlp/)

## Cited by

Cited by

83 papers of our library cite it (of the 140884 whose bibliographies we read); 20 of them have a page here, 63 not yet.

- [ARBERT & MARBERT: Deep Bidirectional Transformers for Arabic](https://md.research.elamur.ai/free/en/papers/abdul-2021-mageed-arbert-marbert-deep-bidirectional-transformers/)
- [AlcLaM: Arabic Dialectal Language Model](https://md.research.elamur.ai/free/en/papers/ahmed-2024-alclam-arabic-dialectal-language-model/)
- [ASAD: A Twitter-based Benchmark Arabic Sentiment Analysis Dataset](https://md.research.elamur.ai/free/en/papers/alharbi-2020-asad-twitter-based-benchmark-arabic/)
- [AraCOVID19-SSD: Arabic COVID-19 Sentiment and Sarcasm Detection Dataset](https://md.research.elamur.ai/free/en/papers/ameur-2021-aracovid19-ssd-arabic-covid-19/)
- [AraBERT: Transformer-based Model for Arabic Language Understanding](https://md.research.elamur.ai/free/en/papers/antoun-2020-arabert-transformer-based-model-arabic/)
- [ArSentD-LEV: A Multi-Topic Corpus for Target-based Sentiment Analysis in Arabic Levantine Tweets](https://md.research.elamur.ai/free/en/papers/baly-2019-arsentd-lev-multi-topic-corpus/)
- [Dhati+: Fine-tuned Large Language Models for Arabic Subjectivity Evaluation](https://md.research.elamur.ai/free/en/papers/bellaouar-2025-dhati-fine-tuned-large-language/)
- [Subjective Sentiment Analysis for Arabic Newswire Comments](https://md.research.elamur.ai/free/en/papers/bessou-2019-subjective-sentiment-analysis-arabic-newswire/)
- [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/)
- [The Interplay of Variant, Size, and Task Type in Arabic Pre-trained Language Models](https://md.research.elamur.ai/free/en/papers/inoue-2021-interplay-variant-size-task-type/)
- [Estimating the Level of Dialectness Predicts Interannotator Agreement in Multi-dialect Arabic Datasets](https://md.research.elamur.ai/free/en/papers/keleg-2024-estimating-level-dialectness-predicts-interannotator/)
- [Preprocessing Solutions for Detection of Sarcasm and Sentiment for Arabic](https://md.research.elamur.ai/free/en/papers/lichouri-2021-preprocessing-solutions-detection-sarcasm-sentiment/)
- [A Deep CNN Architecture with Novel Pooling Layer Applied to Two Sudanese Arabic Sentiment Datasets](https://md.research.elamur.ai/free/en/papers/mhamed-2022-deep-cnn-architecture-novel-pooling/)
- [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/)
- [A review of sentiment analysis research in Arabic language](https://md.research.elamur.ai/free/en/papers/oueslati-2020-review-sentiment-analysis-research-arabic/)
- [Data Augmentation using Transformers and Similarity Measures for Improving Arabic Text Classification](https://md.research.elamur.ai/free/en/papers/refai-2022-data-augmentation-using-transformers-similarity/)
- [WANLP 2021 Shared-Task: Towards Irony and Sentiment Detection in Arabic Tweets using Multi-headed-LSTM-CNN-GRU and MaRBERT](https://md.research.elamur.ai/free/en/papers/salam-2021-wanlp-2021-shared-task-irony/)
- [DeepBlueAI at WANLP-EACL2021 task 2: A Deep Ensemble-based Method for Sarcasm and Sentiment Detection in Arabic](https://md.research.elamur.ai/free/en/papers/song-2021-deepblueai-wanlp-eacl2021-task-2/)
- [Audience Engagement with Arabic Women's Social Empowerment and Wellbeing: A Decadal Corpus](https://md.research.elamur.ai/free/en/papers/zaghouani-2026-audience-engagement-arabic-women-s/)
- [JobArabi: An Arabic Corpus and Analysis of Job Announcements from Social Media](https://md.research.elamur.ai/free/en/papers/zaghouani-2026-jobarabi-arabic-corpus-analysis-job/)

## Cite

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

```bibtex
@inproceedings{nabil2015astd,
  title = {{ASTD: Arabic Sentiment Tweets Dataset}},
  author = {Nabil, Mahmoud and Aly, Mohamed and Atiya, Amir},
  booktitle = {Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing},
  year = {2015},
  doi = {10.18653/v1/d15-1299},
  url = {https://doi.org/10.18653/v1/d15-1299}
}
```

- [← previous Ask Me Anything: Dynamic Memory Networks for Natural Language Processing](https://md.research.elamur.ai/free/en/papers/kumar-2015-ask-me-anything-dynamic-memory/)
- [next → Attention-Based Models for Speech Recognition](https://md.research.elamur.ai/free/en/papers/chorowski-2015-attention-based-models-speech-recognition/)
- [↑ back to papers of 2015](https://md.research.elamur.ai/free/en/papers/2015/)

## References

9 references, 0 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 and Mona T Diab. 2012. Awatif: A multi-genre corpus for modern standard arabic subjectivity and sentiment analysis. In LREC, pages 3907–3914. [find](https://scholar.google.com/scholar?q=Muhammad+Abdul%2DMageed+and+Mona+T+Diab%2E+2012%2E+Awatif%3A+A+multi%2Dgenre+corpus+for+modern+standard+arabic+subjectivity+and+sentiment+analysis%2E+In+LREC%2C+pages+3907%E2%80%933914%2E)
2. Muhammad Abdul-Mageed and Mona Diab. 2014. Sana: A large scale multi-genre, multi-dialect lexicon for arabic subjectivity and sentiment analysis. In Proceedings of the Language Resources and Evaluation Conference (LREC). [find](https://scholar.google.com/scholar?q=Muhammad+Abdul%2DMageed+and+Mona+Diab%2E+2014%2E+Sana%3A+A+large+scale+multi%2Dgenre%2C+multi%2Ddialect+lexicon+for+arabic+subjectivity+and+sentiment+analysis%2E+In+Proceedings+of+the+Language+Resources+and+Evaluatio)
3. Muhammad Abdul-Mageed, Mona Diab, and Sandra Kübler. 2014. Samar: Subjectivity and sentiment analysis for arabic social media. Computer Speech & Language, 28(1):20–37. [find](https://scholar.google.com/scholar?q=Muhammad+Abdul%2DMageed%2C+Mona+Diab%2C+and+Sandra+Ku%CC%88bler%2E+2014%2E+Samar%3A+Subjectivity+and+sentiment+analysis+for+arabic+social+media%2E+Computer+Speech+%26+Language%2C+28%281%29%3A20%E2%80%9337%2E)
4. Mohammed Aly and Amir Atiya. 2013. Labr: Large scale arabic book reviews dataset. In Meetings of the Association for Computational Linguistics (ACL), Sofia, Bulgaria. [find](https://scholar.google.com/scholar?q=Mohammed+Aly+and+Amir+Atiya%2E+2013%2E+Labr%3A+Large+scale+arabic+book+reviews+dataset%2E+In+Meetings+of+the+Association+for+Computational+Linguistics+%28ACL%29%2C+Sofia%2C+Bulgaria%2E)
5. 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. Springer. [find](https://scholar.google.com/scholar?q=Hady+ElSahar+and+Samhaa+R+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%9334%2E+Springer%2E)
6. Hossam S Ibrahim, Sherif M Abdou, and Mervat Gheith. 2015. Sentiment analysis for modern standard arabic and colloquial. arXiv preprint arXiv:1505.03105. [↗](https://arxiv.org/abs/1505.03105)
7. Mahmoud Nabil, Mohamed A. Aly, and Amir F. Atiya. 2014. LABR: A large scale arabic book reviews dataset. CoRR, abs/1411.6718. [↗](https://arxiv.org/abs/1411.6718)
8. Eshrag Refaee and Verena Rieser. 2014. An arabic twitter corpus for subjectivity and sentiment analysis. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC 14), Reykjavik, Iceland, may. European Language Resources Association (ELRA). [find](https://scholar.google.com/scholar?q=Eshrag+Refaee+and+Verena+Rieser%2E+2014%2E+An+arabic+twitter+corpus+for+subjectivity+and+sentiment+analysis%2E+In+Proceedings+of+the+Ninth+International+Conference+on+Language+Resources+and+Evaluation+%28LREC)
9. Mohammed Rushdi-Saleh, M Teresa Martı́n-Valdivia, L Alfonso Ureña-López, and José M Perea-Ortega. 2011. Oca: Opinion corpus for arabic. Journal of the American Society for Information Science and Technology, 62(10):2045–2054, October. [find](https://scholar.google.com/scholar?q=Mohammed+Rushdi%2DSaleh%2C+M+Teresa+Mart%C4%B1%CC%81n%2DValdivia%2C+L+Alfonso+Uren%CC%83a%2DLo%CC%81pez%2C+and+Jose%CC%81+M+Perea%2DOrtega%2E+2011%2E+Oca%3A+Opinion+corpus+for+arabic%2E+Journal+of+the+American+Society+for+Information+Science+and+T)