---
title: "Word Embeddings and Convolutional Neural Network for Arabic"
description: "The authors crawled roughly 10 billion words of Arabic web text, cleaned and deduplicated it, and… 27 references, cited by 22 papers here."
authors: ["Abdelghani Dahou", "Shengwu Xiong", "Junwei Zhou", "Mohamed Houcine Haddoud", "Pengfei Duan"]
year: 2016
venue: "Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers"
source: "https://aclanthology.org/C16-1228/"
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/dahou-2016-word-embeddings-convolutional-neural-network/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/dahou-2016-word-embeddings-convolutional-neural-network/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2016](https://md.research.elamur.ai/free/en/papers/2016/page/3/) / Word Embeddings and Convolutional Neural Network for Arabic Sentiment Classification

# Word Embeddings and Convolutional Neural Network for Arabic Sentiment Classification

**Authors:** Abdelghani Dahou, Shengwu Xiong, Junwei Zhou, Mohamed Houcine Haddoud and Pengfei Duan · **Year:** 2016 · **Venue:** [Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers](https://md.research.elamur.ai/free/en/venues/coling/) ([COLING 2016](https://md.research.elamur.ai/free/en/venues/coling/2016/)) · **Publisher:** [International Committee on Computational Linguistics (ICCL)](https://md.research.elamur.ai/free/en/publishers/international-committee-on-computational-linguistics-iccl/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/C16-1228/) · [PDF](https://aclanthology.org/C16-1228.pdf) · [OpenAlex](https://openalex.org/W2574075017)

**Citations:** 155 citations (OpenAlex, 2026-10-14) · [22](https://md.research.elamur.ai/free/en/papers/dahou-2016-word-embeddings-convolutional-neural-network/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
The paper addresses sentiment polarity detection for Arabic reviews and social media posts. This is hard because Arabic morphology is rich, and much online text is written in dialects rather than Modern Standard Arabic (MSA). The authors crawled roughly 10 billion words of Arabic web text, cleaned and deduplicated it, and used a 3.4-billion-word portion to train word2vec embeddings. They compared CBOW and Skip-Gram at several vector sizes, after first merging multiword expressions into phrase tokens with SegPhrase. Embedding quality was checked in two ways: intrinsically with word analogy questions, and extrinsically by training a single-layer CNN for sentiment on top of the vectors. The CNN was tested on nine datasets, each in a balanced and an unbalanced version, and beat earlier SVM-based results on most of them. Pre-trained initialization (the CBOW58 vectors) gave a clear accuracy gain. The work matters because it shows that a large self-built corpus can compensate for scarce Arabic resources, and the authors released both the code and the vectors.

## Key points
- **Corpus pipeline:**
  - 10,421 seed URLs came from Bing queries generated by BootCAT, which was fed the 2,000 most frequent words with stop words removed.
  - SpiderLing crawled from these seeds.
  - jusText stripped boilerplate.
  - Onion removed near-duplicate paragraphs at a 0.5 threshold.
- **Phrase mining:** A simple bigram-count phrase scorer made phrases about 47.42% of the vocabulary and 16.18% of tokens. Because raw frequency could not separate good phrases from bad ones, the authors switched to SegPhrase, guided by two label sets: 94,544 positive and 121,127 negative.
- **Embedding settings:** word2vec with CBOW (window 5) and Skip-Gram (window 10), dimensions 100, 200 and 300, 10 negative samples, frequency threshold 10, and 3 iterations.
- **Sentiment datasets:** LABR, ASTD, Gold-Standard, an Arabic Twitter set, and the five ElSahar & El-Beltagy domains (HTL, ATT, RES, MOV, PROD).
- **Selected results:**
  - LABR reached 89.6% (unbalanced) versus 78.3% for Linear SVM.
  - HTL reached 91.7% versus 87.6%.
  - The CNN was weaker on RES (78.5 vs 83.6).
  - It was also weaker on Gold-Standard (75.8 vs 87.74), partly because the freely available subset of that dataset is smaller.

## Interesting ideas
- Each sentiment dataset is evaluated twice, once trimmed to equal class sizes and once with all data at natural skew, which separates the effect of label imbalance from the effect of more data.
- The authors found that frequency-based bigram merging produces many junk phrases, so they replaced it with a phrase miner trained on curated lists of good and bad phrases.
- A web corpus is bootstrapped from nothing but a frequent-word list, which yields search seeds, a crawl, boilerplate removal and paragraph-level near-duplicate filtering. This gives a reproducible recipe for any language or domain.
- Embedding quality is judged both intrinsically (analogy questions) and extrinsically (downstream classifier accuracy), so a resource is valued for what it enables, not only for internal metrics.
- Orthographic variants of the same letter (several Alef forms, Teh Marbuta) are normalized before training, so spelling noise does not split one word into many vectors.

## For Mova

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

- **I1** (learn): Evaluation reports could pair every test set with a class-balanced variant, so a gain from skewed data is not mistaken for real understanding.
- **I2** (tokens): Multiword-expression detection (names, fixed compounds) could use a small positive and negative seed list instead of pure co-occurrence counts, keeping each merge decision traceable.
- **I3** (learn): A similar seed-to-clean-text pipeline could gather domain text (school textbooks, fables) for silver parsing, and paragraph dedup would stop repeated pages from inflating agreement statistics.
- **I4** (world): Compiled knowledge such as verb classes or ontology links could be scored both by direct probes and by the answer accuracy it produces downstream.
- **I5** (tokens): The tokenizer could apply an explicit, logged normalization layer for variant characters (quote styles, ligatures), keeping the original form alongside the normalized lemma.

*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

- [Wuhan University of Technology](https://md.research.elamur.ai/free/en/universities/wuhan-university-of-technology/) — University, Wuhan, China · 6 works · 261 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works

🧭 Topics (2): [😊 Sentiment & emotion](https://md.research.elamur.ai/free/en/topics/sentiment-emotion/) · 1,035 works · [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 3,066 works

## Cite

Cite as: Abdelghani Dahou, Shengwu Xiong, Junwei Zhou, Mohamed Houcine Haddoud and Pengfei Duan. 2016. Word Embeddings and Convolutional Neural Network for Arabic Sentiment Classification. Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers.

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

```bibtex
@inproceedings{dahou2016word,
  title = {{Word Embeddings and Convolutional Neural Network for Arabic Sentiment Classification}},
  author = {Abdelghani Dahou, Shengwu Xiong, Junwei Zhou, Mohamed Houcine Haddoud and Pengfei Duan},
  booktitle = {Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers},
  year = {2016},
  url = {https://aclanthology.org/C16-1228/}
}
```

- [← previous Results of the WMT16 Metrics Shared Task](https://md.research.elamur.ai/free/en/papers/bojar-2016-results-wmt16-metrics-shared-task/)
- [next → Who did What: A Large-Scale Person-Centered Cloze Dataset](https://md.research.elamur.ai/free/en/papers/onishi-2016-who-did-what-large-scale/)
- [↑ back to papers of 2016](https://md.research.elamur.ai/free/en/papers/2016/page/3/)

## References

27 references, 12 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, Mona T Diab, and Mohammed Korayem. 2011. [Subjectivity and sentiment analysis of modern standard arabic](https://md.research.elamur.ai/free/en/papers/mageed-2011-subjectivity-sentiment-analysis-modern-standard/). In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies: short papers-Volume 2, pages 587–591.
2. 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. [↗](https://doi.org/10.1016/j.csl.2013.03.001)
3. Nawaf A Abdulla, Nizar A Ahmed, Mohammad A Shehab, and Mahmoud Al-Ayyoub. 2013. Arabic sentiment analysis: Lexicon-based and corpus-based. In 2013 IEEE Jordan Conference on Applied Electrical Engineering and Computing Technologies (AEECT), pages 1–6. [find](https://scholar.google.com/scholar?q=Nawaf+A+Abdulla%2C+Nizar+A+Ahmed%2C+Mohammad+A+Shehab%2C+and+Mahmoud+Al%2DAyyoub%2E+2013%2E+Arabic+sentiment+analysis%3A+Lexicon%2Dbased+and+corpus%2Dbased%2E+In+2013+IEEE+Jordan+Conference+on+Applied+Electrical+Engineer)
4. Mohamed A Aly and Amir F Atiya. 2013. [Labr: A large scale arabic book reviews dataset](https://md.research.elamur.ai/free/en/papers/aly-2013-labr-large-scale-arabic-book/). In ACL (2), pages 494–498. [↗](https://doi.org/10.13140/2.1.3960.5761)
5. Emil St. Chifu, Tiberiu St. Letia, and Viorica R. Chifu. 2015. Unsupervised aspect level sentiment analysis using self-organizing maps. In 2015 17th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing (SYNASC), pages 468–475. [find](https://scholar.google.com/scholar?q=Emil+St%2E+Chifu%2C+Tiberiu+St%2E+Letia%2C+and+Viorica+R%2E+Chifu%2E+2015%2E+Unsupervised+aspect+level+sentiment+analysis+using+self%2Dorganizing+maps%2E+In+2015+17th+International+Symposium+on+Symbolic+and+Numeric+Alg)
6. Ronan Collobert, Jason Weston, Léon Bottou, Michael Karlen, Koray Kavukcuoglu, and Pavel Kuksa. 2011. [Natural language processing (almost) from scratch](https://md.research.elamur.ai/free/en/papers/collobert-2011-natural-language-processing-almost-scratch/). The Journal of Machine Learning Research, 12:2493–2537.
7. John Duchi, Elad Hazan, and Yoram Singer. 2011. Adaptive subgradient methods for online learning and stochastic optimization. The Journal of Machine Learning Research, 12:2121–2159. [find](https://scholar.google.com/scholar?q=John+Duchi%2C+Elad+Hazan%2C+and+Yoram+Singer%2E+2011%2E+Adaptive+subgradient+methods+for+online+learning+and+stochastic+optimization%2E+The+Journal+of+Machine+Learning+Research%2C+12%3A2121%E2%80%932159%2E)
8. Alaa El-Halees. 2011. Arabic opinion mining using combined classification approach. pages 1–8. [find](https://scholar.google.com/scholar?q=Alaa+El%2DHalees%2E+2011%2E+Arabic+opinion+mining+using+combined+classification+approach%2E+pages+1%E2%80%938%2E)
9. 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. [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)
10. Nikos Engonopoulos, Angeliki Lazaridou, Georgios Paliouras, and Konstantinos Chandrinos. 2011. Els: a word-level method for entity-level sentiment analysis. In Proceedings of the International Conference on Web Intelligence, Mining and Semantics, pages 1–9. [find](https://scholar.google.com/scholar?q=Nikos+Engonopoulos%2C+Angeliki+Lazaridou%2C+Georgios+Paliouras%2C+and+Konstantinos+Chandrinos%2E+2011%2E+Els%3A+a+word%2Dlevel+method+for+entity%2Dlevel+sentiment+analysis%2E+In+Proceedings+of+the+International+Confere)
11. Noura Farra, Elie Challita, Rawad Abou Assi, and Hazem Hajj. 2010. Sentence-level and document-level sentiment mining for arabic texts. In Data Mining Workshops (ICDMW), 2010 IEEE International Conference on, pages 1114–1119. [find](https://scholar.google.com/scholar?q=Noura+Farra%2C+Elie+Challita%2C+Rawad+Abou+Assi%2C+and+Hazem+Hajj%2E+2010%2E+Sentence%2Dlevel+and+document%2Dlevel+sentiment+mining+for+arabic+texts%2E+In+Data+Mining+Workshops+%28ICDMW%29%2C+2010+IEEE+International+Confer)
12. Nizar Y Habash. 2010. Introduction to arabic natural language processing. Synthesis Lectures on Human Language Technologies, 3(1):1–187. [↗](https://doi.org/10.1007/978-3-031-02139-8)
13. Bassam Hammo, Hani Abu-Salem, and Steven Lytinen. 2002. [Qarab: A question answering system to support the arabic language](https://md.research.elamur.ai/free/en/papers/hammo-2002-qarab-question-answering-system-support/). In Proceedings of the ACL-02 workshop on Computational approaches to semitic languages, pages 1–11.
14. Geoffrey E Hinton, Nitish Srivastava, Alex Krizhevsky, Ilya Sutskever, and Ruslan R Salakhutdinov. 2012. Improving neural networks by preventing co-adaptation of feature detectors. arXiv preprint arXiv:1207.0580, pages 1–18. [↗](https://arxiv.org/abs/1207.0580)
15. Mohit Iyyer, Peter Enns, Jordan Boyd-Graber, and Philip Resnik. 2014. [Political ideology detection using recursive neural networks](https://md.research.elamur.ai/free/en/papers/iyyer-2014-political-ideology-detection-recursive-neural/). In Association for Computational Linguistics, pages 1113–1122.
16. Nal Kalchbrenner, Edward Grefenstette, and Phil Blunsom. 2014. [A convolutional neural network for modelling sentences](https://md.research.elamur.ai/free/en/papers/kalchbrenner-2014-convolutional-neural-network-modelling-sentences/). arXiv preprint arXiv:1404.2188, pages 655–665. [↗](https://arxiv.org/abs/1404.2188)
17. Yoon Kim. 2014. [Convolutional neural networks for sentence classification](https://md.research.elamur.ai/free/en/papers/kim-2014-convolutional-neural-networks-sentence-classification/). arXiv preprint arXiv:1408.5882, pages 1–6. [↗](https://arxiv.org/abs/1408.5882)
18. Jialu Liu, Jingbo Shang, Chi Wang, Xiang Ren, and Jiawei Han. 2015. Mining quality phrases from massive text corpora. In Proceedings of the 2015 ACM SIGMOD International Conference on Management of Data, pages 1729–1744. [find](https://scholar.google.com/scholar?q=Jialu+Liu%2C+Jingbo+Shang%2C+Chi+Wang%2C+Xiang+Ren%2C+and+Jiawei+Han%2E+2015%2E+Mining+quality+phrases+from+massive+text+corpora%2E+In+Proceedings+of+the+2015+ACM+SIGMOD+International+Conference+on+Management+of+Da)
19. Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg S Corrado, and Jeff Dean. 2013. [Distributed representations of words and phrases and their compositionality](https://md.research.elamur.ai/free/en/papers/mikolov-2013-distributed-representations-words-phrases-their/). In Advances in neural information processing systems, pages 3111–3119. [↗](https://doi.org/10.48550/arxiv.1310.4546)
20. Mahmoud Nabil, Mohamed Aly, and Amir F Atiya. 2015. [Astd: Arabic sentiment tweets dataset](https://md.research.elamur.ai/free/en/papers/nabil-2015-astd-arabic-sentiment-tweets-dataset/). In EMNLP, pages 2515–2519. [↗](https://doi.org/10.60692/3ee2g-zjt73)
21. Jeffrey Pennington, Richard Socher, and Christopher D Manning. 2014. [Glove: Global vectors for word representation](https://md.research.elamur.ai/free/en/papers/pennington-2014-glove-global-vectors-word-representation/). In EMNLP, volume 14, pages 1532–1543. [↗](https://doi.org/10.3115/v1/d14-1162)
22. Jan Pomikálek. 2011. Removing boilerplate and duplicate content from web corpora. Disertacnı práce, Masarykova univerzita, Fakulta informatiky. [find](https://scholar.google.com/scholar?q=Jan+Pomika%CC%81lek%2E+2011%2E+Removing+boilerplate+and+duplicate+content+from+web+corpora%2E+Disertacn%C4%B1+pra%CC%81ce%2C+Masarykova+univerzita%2C+Fakulta+informatiky%2E)
23. Eshrag Refaee and Verena Rieser. 2014. [An arabic twitter corpus for subjectivity and sentiment analysis](https://md.research.elamur.ai/free/en/papers/refaee-2014-arabic-twitter-corpus-subjectivity-sentiment/). In LREC, pages 2268–2273.
24. Richard Socher, Jeffrey Pennington, Eric H Huang, Andrew Y Ng, and Christopher D Manning. 2011. [Semi-supervised recursive autoencoders for predicting sentiment distributions](https://md.research.elamur.ai/free/en/papers/socher-2011-semi-supervised-recursive-autoencoders-predicting/). In EMNLP, pages 151–161.
25. Vı́t Suchomel, Jan Pomikálek, et al. 2012. Efficient web crawling for large text corpora. In Proceedings of the seventh Web as Corpus Workshop (WAC7), pages 39–43. [find](https://scholar.google.com/scholar?q=V%C4%B1%CC%81t+Suchomel%2C+Jan+Pomika%CC%81lek%2C+et+al%2E+2012%2E+Efficient+web+crawling+for+large+text+corpora%2E+In+Proceedings+of+the+seventh+Web+as+Corpus+Workshop+%28WAC7%29%2C+pages+39%E2%80%9343%2E)
26. Ainur Yessenalina, Yisong Yue, and Claire Cardie. 2010. Multi-level structured models for document-level sentiment classification. In EMNLP, pages 1046–1056. [find](https://scholar.google.com/scholar?q=Ainur+Yessenalina%2C+Yisong+Yue%2C+and+Claire+Cardie%2E+2010%2E+Multi%2Dlevel+structured+models+for+document%2Dlevel+sentiment+classification%2E+In+EMNLP%2C+pages+1046%E2%80%931056%2E)
27. Mohamed A Zahran, Ahmed Magooda, Ashraf Y Mahgoub, Hazem Raafat, Mohsen Rashwan, and Amir Atyia. 2015. Word representations in vector space and their applications for arabic. In Computational Linguistics and Intelligent Text Processing, pages 430–443. [find](https://scholar.google.com/scholar?q=Mohamed+A+Zahran%2C+Ahmed+Magooda%2C+Ashraf+Y+Mahgoub%2C+Hazem+Raafat%2C+Mohsen+Rashwan%2C+and+Amir+Atyia%2E+2015%2E+Word+representations+in+vector+space+and+their+applications+for+arabic%2E+In+Computational+Linguist)

*Updated 2 November 2026*