---
title: "Empirical Evaluation of Leveraging Named Entities for Arabic"
description: "The paper asks whether named entities (people, places, organizations) can serve as… 10.7% F1. 21 references, cited by 1 paper here."
authors: ["Hala Mulki", "Hatem Haddad", "Mourad Gridach", "Ismail Babaoglu"]
year: 2019
venue: "arXiv:1904.10195"
source: "https://arxiv.org/abs/1904.10195"
doi: "10.48550/arxiv.1904.10195"
arxiv: "1904.10195"
licence: "arXiv — non-exclusive license to distribute"
licence_url: "http://arxiv.org/licenses/nonexclusive-distrib/1.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/mulki-2019-empirical-evaluation-leveraging-named-entities/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/mulki-2019-empirical-evaluation-leveraging-named-entities/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2019](https://md.research.elamur.ai/free/en/papers/2019/page/7/) / Empirical Evaluation of Leveraging Named Entities for Arabic Sentiment Analysis

# Empirical Evaluation of Leveraging Named Entities for Arabic Sentiment Analysis

**Authors:** [Hala Mulki](https://md.research.elamur.ai/free/en/authors/hala-mulki/), [Hatem Haddad](https://md.research.elamur.ai/free/en/authors/hatem-haddad/), [Mourad Gridach](https://md.research.elamur.ai/free/en/authors/mourad-gridach/) and [Ismail Babaoglu](https://md.research.elamur.ai/free/en/authors/ismail-babaoglu/) · **Year:** 2019 · **Venue:** [arXiv:1904.10195](https://md.research.elamur.ai/free/en/venues/the-international-arab-journal-of-information-technology/) ([The International Arab Journal of Information Technology 2019](https://md.research.elamur.ai/free/en/venues/the-international-arab-journal-of-information-technology/2019/)) · **Publisher:** [Zarqa University](https://md.research.elamur.ai/free/en/publishers/zarqa-university/) · **Licence:** [arXiv — non-exclusive license to distribute](http://arxiv.org/licenses/nonexclusive-distrib/1.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/1904.10195) · [PDF](https://arxiv.org/pdf/1904.10195) · [DOI 10.48550/arxiv.1904.10195](https://doi.org/10.48550/arxiv.1904.10195) · [arXiv 1904.10195](https://arxiv.org/abs/1904.10195) · [OpenAlex](https://openalex.org/W2941304839)

**Citations:** 5 citations (OpenAlex, 2026-10-14) · [1](https://md.research.elamur.ai/free/en/papers/mulki-2019-empirical-evaluation-leveraging-named-entities/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
The paper asks whether named entities (people, places, organizations) can serve as sentiment signals in Arabic social-media sentiment analysis, rather than being ignored or stripped out. In Arabic, many person names are identical to adjectives, which confuses classifiers. The authors first extract entities with an existing BiLSTM-CRF Arabic NER system for social media. They then give each entity a polarity based on whether it appears mostly in positive or mostly in negative posts. In the corpus, every entity is replaced with a PosNE or NegNE tag. They test this inside their Tw-StAR framework, which has a supervised variant and a lexicon-based variant, both using n-gram features. The four multi-dialectal datasets are JEG, TAC, TEC and TSAC. The entity tags did not help the supervised model and slightly hurt it on TAC. They consistently raised F1 for the lexicon-based model, which beat most baselines on the Tunisian sets. The work shows that entity-level priors pay off mainly when the downstream combiner ignores context.

## Key points
- NER uses the deep model of [10]: word embeddings (dim 100) plus character embeddings (dim 25), a BiLSTM and a CRF layer. It reported an F-measure of 85.71% on a Twitter NER set.
- Entity polarity is set by majority vote over the labels of the posts that mention the entity. The approach is inspired by the Straight Forward Sum (SFS) lexicon scoring.
- Lexicon-based Tw-StAR+NEs with uni+bi n-grams reached these F1 scores (without NEs in brackets):
  - JEG: 68.9% (66.3%)
  - TAC: 70.6% (66.6%)
  - TEC: 65.0% (59.8%)
  - TSAC: 82.8% (81.8%)
- In the supervised model, results with and without NEs were roughly equal, except TAC, where F1 dropped by 1.3%. On JEG, the lexicon model stayed 10.7% F1 below a word2vec-based baseline.
- A larger number of annotated entities did not predict a larger gain: JEG improved by +2.6 F1 and TEC by +5.2. The authors attribute gains to corpus consistency, meaning that training and test data share the same entities.

## Interesting ideas
- Give each entity a polarity from the labels of the sentences that mention it, aggregated over a time-bounded corpus; this captures a period-specific public attitude without any hand-built entity lexicon.
- Replace a recognized entity with a typed placeholder token, so a name that is spelled like an ordinary word no longer triggers that word's lexical meaning.
- A prior computed without context helped the context-free summing model but hurt the n-gram learner. The context-free entity tag clashed with contexts the learner had linked to the opposite sentiment.
- The evaluation runs an ablation: an identical pipeline is run once with entity tags and once with entities as plain tokens, across several dialect datasets.
- The authors observe that the overlap of entities between training and test data matters more than how many entities there are.

## For Mova

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

- **I1** (world): entities in the world of a text could carry an accumulated evaluative stance built from the frames they take part in, giving an explainable "how is X regarded here" attribute.
- **I2** (tokens): the tagger and lemmatizer could mark proper-name spans before lexicon lookups, so names such as Hope or Grace never pull in adjective or noun senses downstream.
- **I3** (learn): before adding any global prior feature, test it separately against context-sensitive and context-free consumers, because the same feature can help one and mislead the other.
- **I4** (learn): every new world-level feature could ship with a matched on/off run as a built-in control.
- **I5** (answer): when measuring readers on fables or tests, report entity overlap between development and test material, so that gains from memorized entities are not mistaken for understanding.

*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

- [University of Colorado Boulder](https://md.research.elamur.ai/free/en/universities/university-of-colorado-boulder/) — University, Boulder, United States · 69 works · 51,069 citations
- [Manouba University](https://md.research.elamur.ai/free/en/universities/manouba-university/) — University, Manouba, Tunisia · 5 works · 266 citations
- [Konya Technical University](https://md.research.elamur.ai/free/en/universities/konya-technical-university/) — University, Karatay, Türkiye · 2 works · 262 citations

📚 Shelves (4): [🖥️ 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 · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (3): [😊 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 · [📏 Evaluation & benchmarks](https://md.research.elamur.ai/free/en/topics/evaluation-benchmarks/) · 1,220 works

## Cite

Cite as: Hala Mulki, Hatem Haddad, Mourad Gridach and Ismail Babaoglu. 2019. Empirical Evaluation of Leveraging Named Entities for Arabic Sentiment Analysis. arXiv:1904.10195.

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

```bibtex
@misc{mulki2019empirical,
  title = {{Empirical Evaluation of Leveraging Named Entities for Arabic Sentiment Analysis}},
  author = {Hala Mulki, Hatem Haddad, Mourad Gridach and Ismail Babaoglu},
  year = {2019},
  doi = {10.48550/arxiv.1904.10195},
  eprint = {1904.10195},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1904.10195}
}
```

- [← previous Constrained Sequence-to-sequence Semitic Root Extraction for Enriching Word Embeddings](https://md.research.elamur.ai/free/en/papers/kishky-2019-constrained-sequence-sequence-semitic-root/)
- [next → Experiments in Cuneiform Language Identification](https://md.research.elamur.ai/free/en/papers/paetzold-2019-experiments-cuneiform-language-identification/)
- [↑ back to papers of 2019](https://md.research.elamur.ai/free/en/papers/2019/page/7/)

## References

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

1. A. Aziz A., and Lixin T., “Word embeddings for Arabic sentiment analysis”, IEEE International Conference on Big Data (Big Data), pp. 3820-3825, 2016. [find](https://scholar.google.com/scholar?q=A%2E+Aziz+A%2E%2C+and+Lixin+T%2E%2C+%E2%80%9CWord+embeddings+for+Arabic+sentiment+analysis%E2%80%9D%2C+IEEE+International+Conference+on+Big+Data+%28Big+Data%29%2C+pp%2E+3820%2D3825%2C+2016%2E)
2. Chen D., and Christopher M., “[A fast and accurate dependency parser using neural networks](https://md.research.elamur.ai/free/en/papers/chen-2014-fast-accurate-dependency-parser-neural/)”, Proceedings of 2014 conference on empirical methods in natural language processing (EMNLP), pp. 740-750.
3. Kareem D., “[Named entity recognition using cross-lingual resources: Arabic as an example](https://md.research.elamur.ai/free/en/papers/darwish-2013-named-entity-recognition-cross-lingual/)”, Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), vol. 1, pp. 1558-1567, 2013.
4. Karim S. et al., “Tunisian Dialect and Modern Standard Arabic Dataset for Sentiment Analysis: Tunisian Election Context”, To appear in IEEE proceedings of ACLing 2016. [find](https://scholar.google.com/scholar?q=Karim+S%2E+et+al%2E%2C+%E2%80%9CTunisian+Dialect+and+Modern+Standard+Arabic+Dataset+for+Sentiment+Analysis%3A+Tunisian+Election+Context%E2%80%9D%2C+To+appear+in+IEEE+proceedings+of+ACLing+2016%2E)
5. Mohamed Z., et al., “Word representations in vector space and their applications for Arabic”, International Conference on Intelligent Text Processing and Computational Linguistics, pp. 430-443, 2015. [find](https://scholar.google.com/scholar?q=Mohamed+Z%2E%2C+et+al%2E%2C+%E2%80%9CWord+representations+in+vector+space+and+their+applications+for+Arabic%E2%80%9D%2C+International+Conference+on+Intelligent+Text+Processing+and+Computational+Linguistics%2C+pp%2E+430%2D443%2C+2015%2E)
6. Mohammad S., Saif M., and Svetlana K., “Sentiment after translation: A case-study on A rabic social media posts”, Proceedings of the 2015 conference of the North American chapter of the association for computational linguistics: Human language technologies, pp. 767-777, 2015. [find](https://scholar.google.com/scholar?q=Mohammad+S%2E%2C+Saif+M%2E%2C+and+Svetlana+K%2E%2C+%E2%80%9CSentiment+after+translation%3A+A+case%2Dstudy+on+A+rabic+social+media+posts%E2%80%9D%2C+Proceedings+of+the+2015+conference+of+the+North+American+chapter+of+the+association+fo)
7. Mohammad S., Saif M., and Svetlana K., “How translation alters sentiment”, Journal of Artificial Intelligence Research, vol. 55, pp. 95-130, 2016. [↗](https://doi.org/10.1613/jair.4787)
8. Mohammed A., Nawaf A., and Mahmoud A., “An analytical study of Arabic sentiments: Maktoob case study”, 8th International Conference for Internet Technology and Secured Transactions (ICITST), pp. 89-94, 2013. [find](https://scholar.google.com/scholar?q=Mohammed+A%2E%2C+Nawaf+A%2E%2C+and+Mahmoud+A%2E%2C+%E2%80%9CAn+analytical+study+of+Arabic+sentiments%3A+Maktoob+case+study%E2%80%9D%2C+8th+International+Conference+for+Internet+Technology+and+Secured+Transactions+%28ICITST%29%2C+pp%2E+89%2D94)
9. Mourad A., and Karim D., “[Subjectivity and sentiment analysis of modern standard Arabic and Arabic microblogs](https://md.research.elamur.ai/free/en/papers/mourad-2013-subjectivity-sentiment-analysis-modern-standard/)”, Proceedings of the 4th workshop on computational approaches to subjectivity, sentiment and social media analysis, pp. 55-64. 2013.
10. Mourad G., “[Character-Aware Neural Networks for Arabic Named Entity Recognition for Social Media](https://md.research.elamur.ai/free/en/papers/gridach-2016-character-aware-neural-networks-arabic/)”, Proceedings of the 6th Workshop on South and Southeast Asian Natural Language Processing (WSSANLP), pp. 23-32, 2016.
11. Nabil M., Mohamed A., and Amir A., “[Astd: Arabic sentiment tweets dataset](https://md.research.elamur.ai/free/en/papers/nabil-2015-astd-arabic-sentiment-tweets-dataset/)”, Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing (EMNLP),pp. 2515-2519, 2015. [↗](https://doi.org/10.60692/3ee2g-zjt73)
12. Nadia K., Tunisian Arabic Customer's Reviews Processing And Analysis For an Internet Supervision System, Sfax University, 2017. [find](https://scholar.google.com/scholar?q=Nadia+K%2E%2C+Tunisian+Arabic+Customer%27s+Reviews+Processing+And+Analysis+For+an+Internet+Supervision+System%2C+Sfax+University%2C+2017%2E)
13. Najwa E., et al., “Sentiment analysis of colloquial Arabic tweets”, ASE BigData, Social Informatics, PASSAT, BioMedCom Conference, Harvard University, pp. 1-9, 2013. [find](https://scholar.google.com/scholar?q=Najwa+E%2E%2C+et+al%2E%2C+%E2%80%9CSentiment+analysis+of+colloquial+Arabic+tweets%E2%80%9D%2C+ASE+BigData%2C+Social+Informatics%2C+PASSAT%2C+BioMedCom+Conference%2C+Harvard+University%2C+pp%2E+1%2D9%2C+2013%2E)
14. Nawaf A., et al., “Arabic sentiment analysis: Lexicon-based and corpus-based”, IEEE Jordan Conference on Applied Electrical Engineering and Computing Technologies (AEECT), pp. 1-6, 2013. [find](https://scholar.google.com/scholar?q=Nawaf+A%2E%2C+et+al%2E%2C+%E2%80%9CArabic+sentiment+analysis%3A+Lexicon%2Dbased+and+corpus%2Dbased%E2%80%9D%2C+IEEE+Jordan+Conference+on+Applied+Electrical+Engineering+and+Computing+Technologies+%28AEECT%29%2C+pp%2E+1%2D6%2C+2013%2E)
15. Nawaf A., et al., “Towards improving the lexicon-based approach for arabic sentiment analysis”, International Journal of Information Technology and Web Engineering (IJITWE), vol.9, no.3, pp. 55-71, 2014. [find](https://scholar.google.com/scholar?q=Nawaf+A%2E%2C+et+al%2E%2C+%E2%80%9CTowards+improving+the+lexicon%2Dbased+approach+for+arabic+sentiment+analysis%E2%80%9D%2C+International+Journal+of+Information+Technology+and+Web+Engineering+%28IJITWE%29%2C+vol%2E9%2C+no%2E3%2C+pp%2E+55%2D71%2C+20)
16. Quoc L., and Tomas M., “[Distributed representations of sentences and documents](https://md.research.elamur.ai/free/en/papers/le-2014-distributed-representations-sentences-documents/)”, International Conference on Machine Learning, pp. 1188-1196, 2014.
17. Rajesh P., Devasher M., and Vivek Kumar S., “Analytical mapping of opinion mining and sentiment analysis research during 2000–2015”, Information Processing & Management, vol.53, no.1,pp. 122-150, 2017. [find](https://scholar.google.com/scholar?q=Rajesh+P%2E%2C+Devasher+M%2E%2C+and+Vivek+Kumar+S%2E%2C+%E2%80%9CAnalytical+mapping+of+opinion+mining+and+sentiment+analysis+research+during+2000%E2%80%932015%E2%80%9D%2C+Information+Processing+%26+Management%2C+vol%2E53%2C+no%2E1%2Cpp%2E+122%2D150%2C+2017)
18. Salima M., et al., “Sentiment analysis of Tunisian dialects: Linguistic resources and experiments”, Proceedings of the Third Arabic Natural Language Processing Workshop, pp.55-61, 2017. [find](https://scholar.google.com/scholar?q=Salima+M%2E%2C+et+al%2E%2C+%E2%80%9CSentiment+analysis+of+Tunisian+dialects%3A+Linguistic+resources+and+experiments%E2%80%9D%2C+Proceedings+of+the+Third+Arabic+Natural+Language+Processing+Workshop%2C+pp%2E55%2D61%2C+2017%2E)
19. Samhaa E., and Ahmad A., “Open issues in the sentiment analysis of Arabic social media: A case study”, 9th IEEE international conference on Innovations in information technology (iit), pp. 215-220, 2013. [find](https://scholar.google.com/scholar?q=Samhaa+E%2E%2C+and+Ahmad+A%2E%2C+%E2%80%9COpen+issues+in+the+sentiment+analysis+of+Arabic+social+media%3A+A+case+study%E2%80%9D%2C+9th+IEEE+international+conference+on+Innovations+in+information+technology+%28iit%29%2C+pp%2E+215%2D220%2C+20)
20. Samhaa E., et al., “Combining Lexical Features and a Supervised Learning Approach for Arabic Sentiment Analysis”, International Conference on Intelligent Text Processing and Computational Linguistics, pp. 307-319, 2016. [find](https://scholar.google.com/scholar?q=Samhaa+E%2E%2C+et+al%2E%2C+%E2%80%9CCombining+Lexical+Features+and+a+Supervised+Learning+Approach+for+Arabic+Sentiment+Analysis%E2%80%9D%2C+International+Conference+on+Intelligent+Text+Processing+and+Computational+Linguistics%2C)
21. Tomas M., et al., “[Distributed representations of words and phrases and their compositionality](https://md.research.elamur.ai/free/en/papers/mikolov-2013-distributed-representations-words-phrases-their/)”, Advances in neural information processing system, pp. 3111-3119, 2013.

*Updated 1 November 2026*