Elamur Research⌕
☰
home / papers / 2017 / ARAACOM: ARAbic Algerian Corpus for Opinion Mining

ARAACOM: ARAbic Algerian Corpus for Opinion Mining

Paper · 2017 · Zitouni Abdelhafid, Hichem Rahab, Abdelhafid Zitouni et al.

Summary

The paper tackles polarity classification (positive vs. negative) of reader comments posted on Algerian Arabic news websites. Public Arabic sentiment datasets are scarce and mostly built from movie reviews, so the authors built ARAACOM, a small corpus of comments from the Echorouk newspaper site. Articles span news, politics, sport and culture. The site lets readers vote on other comments rather than rate the article, so votes say nothing reliable about polarity. Every comment therefore had to be read and labelled by hand. Off-topic and neutral comments were set aside. The final corpus has 92 Modern Standard Arabic comments. Using RapidMiner, the authors trained SVM and Naïve Bayes with several parameter settings and evaluated them with 10-fold cross-validation. Both models scored well on precision and gained from bigram features, while F-measure stayed modest. Against the established OCA corpus, ARAACOM gave competitive figures, and SVM precision on ARAACOM exceeded OCA's. The work matters as an early domain-specific resource for Algerian news commentary.

Key points

  • ARAACOM contains 92 comments from Echorouk: 60 negative and all 32 available positive ones, a 65.21% / 34.78% split. Before reduction there were 91 negative comments, plus neutral ones.
  • Comments average 36.45 tokens. The authors cite earlier findings that document length affects classification more than dataset size, and use this to justify the small corpus.
  • Practical obstacles in the data:
    • frequent misspellings;
    • comments wholly or partly in French;
    • surprisingly few comments in Algerian dialect;
    • generally low comment volume;
    • many off-topic comments.
  • The experiments used SVM and NB in RapidMiner with 10-fold cross-validation, reporting precision, recall and F-measure. Bigrams improved both classifiers.
  • Planned future work:
    • adding a neutral class;
    • adding other Algerian newspapers;
    • handling dialect and French without manual translation;
    • conditioning comment sentiment on the article's topic.

Interesting ideas

  • I1. On news sites, up/down votes on comments measure agreement with the commenter, not the comment's polarity, so the authors rejected them as labels instead of trusting an easy proxy.
  • I2. Many comments only make sense relative to the article they respond to, so their sentiment is really a stance toward a target rather than a property of the text alone.
  • I3. Off-topic and neutral comments were kept apart as their own category rather than forced into a positive/negative decision.
  • I4. A new, small corpus was benchmarked against an established one (OCA) using the same classifiers and validation scheme, which makes the comparison interpretable despite the size gap.
  • I5. The small corpus was defended by pointing to long documents (about 36 tokens each), on the view that the amount of text per item can matter more than the number of items.

For Mova

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

  • I1 (learn): before treating agreement among witness parsers or LLM teacher tiers as silver truth, check on a gold sample that agreement actually tracks correctness.
  • I2 (frames): attach polarity in prag frames to an explicit target entity or proposition, so evaluations stay grounded and explainable instead of floating as document scores.
  • I3 (answer): readers could get an explicit "not about this question" abstain path, which also gives evaluation a natural negative control.
  • I4 (learn): when adding a new treebank or test set, rerun the same pipeline on a known reference set to separate data effects from model effects.
  • I5 (learn): size and report evaluation sets in tokens or dependency arcs, not only in sentence counts, when judging whether a sample is big enough.

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

🎓Université Constantine 2Constantine · Algeria3 works · 18 citations🎓Abbès Laghrour University of KhenchelaKhenchela · Algeria2 works · 7 citations🎓Université de PoitiersPoitiers · France2 works · 7 citations

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 9,086 worksArtificial Intelligence · 8,889 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 worksPhysics and Astronomy · 25 worksStatistical and Nonlinear Physics · 22 works

🧭 Topics 2Research subjects and languages the work is about.Sentiment & emotion · 608 worksSemitic languages · 1,484 works

Cite

Cite as Zitouni Abdelhafid, Hichem Rahab, Abdelhafid Zitouni and Mahieddine Djoudi. 2017. ARAACOM: ARAbic Algerian Corpus for Opinion Mining. arXiv:2001.08010 · ICCES '17: Proceedings of the International Conference on Computing for Engineering and Sciences, Jul 2017, Istanbul, France. pp.35-39.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.1145/3129186.3129193 · arXiv 2001.08010

@misc{abdelhafid2017araacom,
  title = {{ARAACOM: ARAbic Algerian Corpus for Opinion Mining}},
  author = {Zitouni Abdelhafid, Hichem Rahab, Abdelhafid Zitouni and Mahieddine Djoudi},
  year = {2017},
  doi = {10.1145/3129186.3129193},
  eprint = {2001.08010},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.1145/3129186.3129193}
}

References

26 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Agarwal, B. and Mittal, N. 2016. Prominent Feature Extraction for Sentiment Analysis. ↗
  2. Alotaibi, S.S. 2015. Sentiment Analysis in the Arabic Language Using Machine Learning. PHD thesis, Colorado State University. ↗
  3. Alotaibi, S.S. and Anderson, C.W. 2016. Extending the Knowledge of the Arabic Sentiment Classification Using a Foreign External Lexical Source. International Journal on Natural Language Computing. 5, 3 (2016), 1–11. ↗
  4. Atia, S. and Shaalan, K. 2015. Increasing the accuracy of opinion mining in Arabic. Proceedings - 1st International Conference on Arabic Computational Linguistics: Advances in Arabic Computational Linguistics, ACLing 2015. (2015), 106–113. find
  5. Brahimi, B. et al. 2016. Data and Text Mining Techniques for Classifying Arabic Tweet Polarity. Journal of Digital Information Management. 14, 1 (2016), 15–25. find
  6. Cherif, W. et al. 2015. Towards an Efficient Opinion Measurement in Arabic Comments. Procedia Computer Science. 73, Awict (2015), 122–129. find
  7. Cortes, C. and Vapnik, V. 1995. Support-Vector Networks. Machine Learning. 20, 3 (1995), 273–297. ↗
  8. Duwairi, R.M. 2015. Sentiment analysis for dialectical Arabic. 6th International Conference on Information and Communication Systems (ICICS) (2015), 166–170. find
  9. El-defrawy, M. 2015. Enhancing Root Extractors Using Light Stemmers. 29th Pacific Asia Conference on Language, Information and Computation: (2015), 157– 166. find
  10. El-halees, A. 2011. Arabic Opinion Mining Using Combined Classification Approach. Proceeding The International Arab Conference On Information Technology, Azrqa, Jordan. (2011). find
  11. Farghaly, A. and Shaalan, K. 2009. Arabic natural language processing : Challenges and solutions. ACM Transactions on Asian Language Information Processing. 8, 4 (2009), 1–22. ↗
  12. Habash, N. et al. 2007. On Arabic Transliteration. Arabic Computational Morphology. 15–22. ↗
  13. Huan, L. and Lei, Y. 2005. Toward integrating feature selection algorithms for classification and clustering. Ieee Transactions on Knowledge and Data Engineering. 17, 4 (2005), 491–502. find
  14. Jackson, P. and Moulinier, I. 2002. Natural Language Processing for Online Aplications: Text Retrieval, Extraction and Categorization. John Benjamins Publishing Company. find
  15. Liu, B. 2012. Sentiment Analysis and Opinion Mining. Synthesis Lectures on Human Language Technologies. 5, 1 (2012), 1–167. ↗
  16. Liu, B. 2010. Sentiment Analysis and Subjectivity. Handbook of Natural Language Processing. find
  17. Matthew, P. and Jessica, B. 2010. Introduction to Manual Annotation. Fairview Research. (2010). find
  18. McCallum, A. and Nigam, K. 1998. A Comparison of Event Models for Naive Bayes Text Classification. AAAI/ICML-98 Workshop on Learning for Text Categorization. (1998), 41–48. find
  19. Meftouh, K. et al. 2012. a Study of a Non-Resourced Language : an Algerian Dialect. Proc. 3td International Workshop on Spoken Languages Technologies for Under-resourced Languages (SLTU’12), Cape Town, South Africa, 7-9 May, (2012). find
  20. Mountassir, A. et al. 2013. Sentiment classification on arabic corpora. A preliminary cross-study. Document numérique. 16, 1 (2013), 73–96. find
  21. Rushdi-Saleh, M. et al. 2011. OCA: Opinion corpus for Arabic. Journal of the American Society for Information Science and Technology. 62, 10 (2011), 2045–2054. ↗
  22. Rushdi-Saleh, M. and Martín-Valdivia, M. 2011. Bilingual experiments with an arabic-english corpus for opinion mining. Proceedings of the International Conference Recent Advances in Natural Language Processing 2011. September (2011), 740–745. find
  23. Salloum, S.A. et al. 2017. A Survey of Text Mining in Social Media : Facebook and Twitter Perspectives. Advances in Science, Technology and Engineering Systems Journal. 2, 1 (2017), 127–133. ↗
  24. Sindhu Bhargavi; Sruthi, Namburi; Pavan, Kumar, A.V. 2016. Opinion Mining on Indian Newspaper Quotations. Proceedings - 2009 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology - Workshops, WI-IAT Workshops 2009. 3, Dc (2016), 523–526. find
  25. Zhang, C. et al. 2009. Sentiment analysis of chinese documents: From sentence to document level. Journal of the American Society for Information Science and Technology. 60, 12 (2009), 2474–2487. ↗
  26. Ziani, A. et al. 2013. Détection de polarité d’opinion dans les forums en langues arabe par fusion de plusieurs SVMs. 7, (2013), 17–21. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/abdelhafid-2020-araacom-arabic-algerian-corpus-opinion/