---
title: "SentiArabic: A Sentiment Analyzer for Standard Arabic (2018)"
description: "The author builds SentiArabic, a binary (positive/negative) polarity classifier designed to stay computationally cheap. 76.5% F1. 17 references."
authors: ["Ramy Eskander"]
year: 2018
venue: "Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)"
source: "https://aclanthology.org/L18-1195/"
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/eskander-2018-sentiarabic-sentiment-analyzer-standard-arabic/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/eskander-2018-sentiarabic-sentiment-analyzer-standard-arabic/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2018](https://md.research.elamur.ai/free/en/papers/2018/page/6/) / SentiArabic: A Sentiment Analyzer for Standard Arabic

# SentiArabic: A Sentiment Analyzer for Standard Arabic

**Authors:** [Ramy Eskander](https://md.research.elamur.ai/free/en/authors/ramy-eskander/) · **Year:** 2018 · **Venue:** [Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)](https://md.research.elamur.ai/free/en/venues/lrec/) ([LREC 2018](https://md.research.elamur.ai/free/en/venues/lrec/2018/)) · **Publisher:** [European Language Resources Association (ELRA)](https://md.research.elamur.ai/free/en/publishers/european-language-resources-association-elra/) · **Licence:** [CC BY-NC 4.0 (ELRA, LREC proceedings)](https://creativecommons.org/licenses/by-nc/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/L18-1195/) · [PDF](https://aclanthology.org/L18-1195.pdf) · [OpenAlex](https://openalex.org/W2805157301)

**Citations:** 4 citations (OpenAlex, 2026-10-14)

## Summary
The paper addresses the lack of practical sentiment tools for Standard Arabic, where scarce resources and rich morphology make analysis costly. The author builds SentiArabic, a binary (positive/negative) polarity classifier designed to stay computationally cheap. There are three contributions. First, an improved version of the SLSA lexicon, with hand-corrected entries for frequent lemmas, added entries, and coarsened scores. Second, a new 4,000-sentence news corpus annotated for contextual polarity. Third, the analyzer itself. Instead of running a full morphological tagger, it maps each word to its most frequent lemma and POS in PATB. It then retrieves lexicon polarity scores and decides the sentence label with a simple decision tree over those scores. Negation is handled by swapping positive and negative scores in the context that follows. On the PATB-derived test set of Abdul-Mageed et al. (2011), SentiArabic reaches 76.5% F1. This is 3.0 points above a deep-learning system (Sallab et al., 2015). The result suggests that a carefully curated lexicon plus shallow processing can rival heavier models for morphologically rich languages.

## Key points
- SLSA was created automatically by linking the AraMorph lexicon to SentiWordNet through English glosses. The paper reports roughly 34K lemma-POS entries (both 34,281 and 34,852 appear) and an estimated entry accuracy of 80.1%. Most errors come from questionable SentiWordNet scores.
- The SLSA extension involves three changes:
  - manual correction of entries for the 4,000 most frequent lemmas in PATB Part 3;
  - 220 new lemma-POS entries missing from AraMorph;
  - positive and negative scores rounded up to multiples of 0.25.
- The new corpus has 4,000 news sentences across genres such as politics, sports, religion, and medicine. It is split into SentiTrain (3,200), SentiDev (400), and SentiTest (400). For mixed sentences, the dominant sentiment is chosen as the label.
- Results:

  | Test set | SentiArabic F1 | Without extension F1 | Comparison |
  |---|---|---|---|
  | SentiTest | 80.8% | 67.0% | — |
  | PATB set | 76.5% | 70.8% | Sallab et al. 73.5%; Eskander & Rambow 68.6% |

- Using MADAMIRA instead of the frequency lookup improves F1 by only 0.5 points, which the author judges not worth the extra computation. About 15% of the errors are mixed-polarity sentences where the less dominant sentiment was chosen.

## Interesting ideas
- Replace a full contextual tagger with a lookup of each word form's most frequent treebank analysis, then measure exactly how much accuracy is lost (here only 0.5 F1).
- Round noisy, automatically derived scores into a few coarse bins (steps of 0.25) so the classifier ignores spurious small differences between entries.
- Focus manual correction on the most frequent lemmas first, since those affect the most tokens; here this curation drove most of the gain over the unextended lexicon.
- Model negation as an operation that swaps positive and negative contributions within its scope, rather than as just another feature.
- Run an ablation with and without the resource extension on two different test sets, plus an error breakdown that isolates mixed-polarity cases.

## For Mova

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

- **I1** (tokens): Build a most-frequent-analysis lemma/POS table from UD treebanks as a fast fallback for the en crate. Report its gap against the full tagger so the cost of the cheap path stays explicit.
- **I2** (world): Values imported from WordNet/FrameNet-derived resources could be quantized into a few explainable levels, making downstream rules robust to noise in the import.
- **I3** (learn): Rank lexicon or silver-data items by treebank frequency before sending them to the LLM teacher tiers or human review, maximizing corrected coverage per unit of effort.
- **I4** (frames): In prag, a negation attached in the UD tree could flip the polarity of the frame it governs. Using tree scope instead of linear "following context" would keep the step explainable.
- **I5** (learn): Evaluate each new compiled resource with an on/off ablation across several test sets, and tag error classes, so improvements are attributable rather than anecdotal.

*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

- [Columbia University](https://md.research.elamur.ai/free/en/universities/columbia-university/) — University, New York, United States · 149 works · 48,852 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,086 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,889 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/) · 608 works · [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 1,484 works

## Cite

Cite as: Ramy Eskander. 2018. SentiArabic: A Sentiment Analyzer for Standard Arabic. Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018).

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

```bibtex
@inproceedings{eskander2018sentiarabic,
  title = {{SentiArabic: A Sentiment Analyzer for Standard Arabic}},
  author = {Ramy Eskander},
  booktitle = {Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)},
  year = {2018},
  doi = {10.63317/3matbrdtj2yp},
  url = {https://doi.org/10.63317/3matbrdtj2yp}
}
```

- [← previous Predicting the Argumenthood of English Prepositional Phrases](https://md.research.elamur.ai/free/en/papers/kim-2018-predicting-argumenthood-english-prepositional-phrases/)
- [next → Sprucing up the trees – Error detection in treebanks](https://md.research.elamur.ai/free/en/papers/rehbein-2018-sprucing-up-trees-error-detection/)
- [↑ back to papers of 2018](https://md.research.elamur.ai/free/en/papers/2018/page/6/)

## References

17 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Abdul-Mageed, M., Diab, M., and Korayem, M. (2011). Subjectivity and sentiment analysis of modern standard arabic. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Portland, Oregon. [find](https://scholar.google.com/scholar?q=Abdul%2DMageed%2C+M%2E%2C+Diab%2C+M%2E%2C+and+Korayem%2C+M%2E+%282011%29%2E+Subjectivity+and+sentiment+analysis+of+modern+standard+arabic%2E+In+Proceedings+of+the+49th+Annual+Meeting+of+the+Association+for+Computational+Lingui)
2. Abdul-Mageed, M., Diab, M., and Kubler, S. (2014). Samar: Subjectivity and sentiment analysis for arabic social media. Computer Speech and Language, 28:20–37. [↗](https://doi.org/10.1016/j.csl.2013.03.001)
3. Aly, M. and Atiya, A. (2013). Labr: A large scale arabic book reviews dataset. In Proceedings of the 11th International Workshop on the ACL2 Theorem Prover and Its Applications, Laramie, Wyoming. [↗](https://doi.org/10.13140/2.1.3960.5761)
4. A.M. Alayba, V. Palade, M. E. R. I. (2017). Arabic language sentiment analysis on health services. In Proceedings of the 1st International Workshop on Arabic Script Analysis and Recognition (ASAR), Nancy, France. [find](https://scholar.google.com/scholar?q=A%2EM%2E+Alayba%2C+V%2E+Palade%2C+M%2E+E%2E+R%2E+I%2E+%282017%29%2E+Arabic+language+sentiment+analysis+on+health+services%2E+In+Proceedings+of+the+1st+International+Workshop+on+Arabic+Script+Analysis+and+Recognition+%28ASAR%29%2C+Na)
5. Baccianella, S., Esuli, A., and Sebastiani, F. (2010). [Sentiwordnet 3.0: An enhanced lexical resource for sentiment analysis and opinion mining](https://md.research.elamur.ai/free/en/papers/baccianella-2010-sentiwordnet-3-0-enhanced-lexical/). In Proceedings of the 7th international conference on Language Resources and Evaluation (LREC), Valletta, Malta.
6. T. Buckwalter. (2004). Buckwalter Arabic Morphological Analyzer (Aramorph) Version 2.0. Linguistic Data Consortium (LDC), 2. [find](https://scholar.google.com/scholar?q=T%2E+Buckwalter%2E+%282004%29%2E+Buckwalter+Arabic+Morphological+Analyzer+%28Aramorph%29+Version+2%2E0%2E+Linguistic+Data+Consortium+%28LDC%29%2C+2%2E)
7. El-Beltagy, S. and Ali, A. (2013). Open issues in the sentiment analysis of arabic social media: a case study. In Proceedings of the 9th International Conference on Innovations in Information Technology (IIT), Al Ain, AUE. [find](https://scholar.google.com/scholar?q=El%2DBeltagy%2C+S%2E+and+Ali%2C+A%2E+%282013%29%2E+Open+issues+in+the+sentiment+analysis+of+arabic+social+media%3A+a+case+study%2E+In+Proceedings+of+the+9th+International+Conference+on+Innovations+in+Information+Technolo)
8. El-Makky, N., Nagi, K., El-Ebshihy, A., and Ibrahim, S. (2014). Sentiment analysis of colloquial arabic tweets. In Proceedings of the 3rd ASE International Conference on Social Informatics, Boston, Massachusetts. [find](https://scholar.google.com/scholar?q=El%2DMakky%2C+N%2E%2C+Nagi%2C+K%2E%2C+El%2DEbshihy%2C+A%2E%2C+and+Ibrahim%2C+S%2E+%282014%29%2E+Sentiment+analysis+of+colloquial+arabic+tweets%2E+In+Proceedings+of+the+3rd+ASE+International+Conference+on+Social+Informatics%2C+Boston%2C+Ma)
9. Eskander, R. and Rambow, O. (2015). Slsa: A sentiment lexicon for standard arabic. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, Lisbon, Portugal. [find](https://scholar.google.com/scholar?q=Eskander%2C+R%2E+and+Rambow%2C+O%2E+%282015%29%2E+Slsa%3A+A+sentiment+lexicon+for+standard+arabic%2E+In+Proceedings+of+the+2015+Conference+on+Empirical+Methods+in+Natural+Language+Processing%2C+Lisbon%2C+Portugal%2E)
10. Kiritchenko, S., Mohammad, S., and Salameh, M. (2016). Semeval-2016 task 7: Determining sentiment intensity of english and arabic phrases. In Proceedings of the 2016 International Workshop on Semantic Evaluation (SemEval), San Diego, California. [find](https://scholar.google.com/scholar?q=Kiritchenko%2C+S%2E%2C+Mohammad%2C+S%2E%2C+and+Salameh%2C+M%2E+%282016%29%2E+Semeval%2D2016+task+7%3A+Determining+sentiment+intensity+of+english+and+arabic+phrases%2E+In+Proceedings+of+the+2016+International+Workshop+on+Semantic)
11. Maamouri, M., Bies, A., Buckwalter, T., and Mekki, W. (2004). Arabic treebank: Building a large-scale annotated arabic corpus. In Proceedings of the NEMLAR Conference on Arabic Language Resources and Tools, Cairo, Egypt. [find](https://scholar.google.com/scholar?q=Maamouri%2C+M%2E%2C+Bies%2C+A%2E%2C+Buckwalter%2C+T%2E%2C+and+Mekki%2C+W%2E+%282004%29%2E+Arabic+treebank%3A+Building+a+large%2Dscale+annotated+arabic+corpus%2E+In+Proceedings+of+the+NEMLAR+Conference+on+Arabic+Language+Resources+and+)
12. Mourad, A. and Darwish, K. (2013). Subjectivity and sentiment analysis of modern standard arabic and arabic microblogs. In Proceedings of the 4th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis (WASSA), Atlanta, Georgia. [find](https://scholar.google.com/scholar?q=Mourad%2C+A%2E+and+Darwish%2C+K%2E+%282013%29%2E+Subjectivity+and+sentiment+analysis+of+modern+standard+arabic+and+arabic+microblogs%2E+In+Proceedings+of+the+4th+Workshop+on+Computational+Approaches+to+Subjectivity%2C+)
13. Nabil, M., Aly, M., and Atiya, A. (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 (EMNLP), Lisbon, Portugal. [↗](https://doi.org/10.60692/3ee2g-zjt73)
14. Pasha, A., Al-Badrashiny, M., Diab, M., Kholy, A. E., Eskander, R., Habash, N., Pooleery, M., Rambow, O., and Roth, R. (2014). [Madamira: A fast, comprehensive tool for morphological analysis and disambiguation of arabic](https://md.research.elamur.ai/free/en/papers/pasha-2014-madamira-fast-comprehensive-tool-morphological/). In Proceedings of the 9th international conference on Language Resources and Evaluation (LREC), Reykjavik, Iceland.
15. Refaee, E. and Rieser, V. (2016). ilab-edinburgh at semeval-2016 task 7: A hybrid approach for determining sentiment intensity of arabic twitter phrases. In Proceedings of the 2016 International Workshop on Semantic Evaluation (SemEval), San Diego, California. [find](https://scholar.google.com/scholar?q=Refaee%2C+E%2E+and+Rieser%2C+V%2E+%282016%29%2E+ilab%2Dedinburgh+at+semeval%2D2016+task+7%3A+A+hybrid+approach+for+determining+sentiment+intensity+of+arabic+twitter+phrases%2E+In+Proceedings+of+the+2016+International+Works)
16. Sallab, A. A., Baly, R., Badaro, G., Hajj, H., and W. El-Hajj, K. S. (2015). Deep learning models for sentiment analysis in arabic. In Proceedings of the EMNLP 2015 Workshop on Arabic Natural Language Processing (ANLP), Lisbon, Portugal. [find](https://scholar.google.com/scholar?q=Sallab%2C+A%2E+A%2E%2C+Baly%2C+R%2E%2C+Badaro%2C+G%2E%2C+Hajj%2C+H%2E%2C+and+W%2E+El%2DHajj%2C+K%2E+S%2E+%282015%29%2E+Deep+learning+models+for+sentiment+analysis+in+arabic%2E+In+Proceedings+of+the+EMNLP+2015+Workshop+on+Arabic+Natural+Language)
17. Wilson, T., Wiebe, J., and Hoffmann, P. (2005). [Recognizing contextual polarity in phrase-level sentiment analysis](https://md.research.elamur.ai/free/en/papers/wilson-2005-recognizing-contextual-polarity-phrase-level/). In Proceedings of the 1st Conference on Empirical Methods in Natural Language Processing (EMNLP), Vancouver, Canada.

*Updated 24 October 2026*