Elamur Research⌕
☰
home / papers / 2014 / A Large Scale Arabic Sentiment Lexicon for Arabic Opinion Mining

A Large Scale Arabic Sentiment Lexicon for Arabic Opinion Mining

Paper · 2014 · Gilbert Badaro, Ramy Baly, Hazem Hajj et al.

Summary

Badaro and colleagues tackle the lack of a large, public Arabic sentiment lexicon that fits Arabic's rich morphology. Earlier Arabic resources were small, mixed lemmas with inflected forms, or were never released. The authors build ArSenL, a lemma-level lexicon. Each entry is a SAMA lemma carrying positive, negative and objective scores taken from English SentiWordNet (ESWN). They compare two construction routes:

  • ArSenL-AWN: follows Arabic WordNet offsets into ESWN, then aligns AWN lemmas to SAMA.
  • ArSenL-Eng: links SAMA lemmas to ESWN through their English glosses, which needs only a dictionary.

The final resource is the union of both. They evaluate it extrinsically with SVM classifiers for subjectivity and polarity on a manually annotated part of the Penn Arabic Treebank. The comparison includes a majority baseline and the hand-built adjective lexicon SIFAAT. The gloss-based route beats the wordnet route, and the union beats everything, including SIFAAT. This matters because the cheap dictionary route offers a template for Arabic dialects and other low-resource languages.

Key points

  • ArSenL covers about 28,760 lemmas and 157,969 synset entries, linked to SAMA 3.1 lemmas so it works after lemmatization (e.g., with MADA).
  • The AWN route was narrow:
    • 9,635 synsets (6,967 lemmas) reached ESWN 3.0 through version sense maps 2.0→2.1→3.0.
    • WordNet has no adjective sense maps, so only nouns and verbs survived.
    • The 1,695 multi-word lemmas (24%) were dropped.
  • AWN-to-SAMA alignment was staged and ended with 5,002 AWN lemmas tied to 4,507 SAMA lemmas, only 12% of SAMA:
    • exact match gave 1,736 lemmas;
    • normalization rules added 1,927;
    • analyzer plus edit-distance backoff added 1,094.
  • Lemma coverage on the evaluation corpus:
LexiconCoverage
Union89.9%
Eng88.8%
AWN56.6%
SIFAAT32.1%
  • Average positive/negative F1:
LexiconAvg. F1 (Pos/Neg)
Union64.5
Eng62.4
SIFAAT59.2
AWN53.9
Majority baseline36.9
  • The best classifier setup used a Gaussian RBF kernel and sentence-length normalization, without confidence weighting.

Interesting ideas

  • I1. Synset offsets from different WordNet versions are chained through sense-map files, with POS tags rechecked at each hop to catch reused offsets. This is a cheap, verifiable way to carry annotations across resource versions.
  • I2. Lemma alignment between two resources runs as an ordered cascade: exact match, then orthographic and lemma-form normalization, then analyzer readings ranked by edit distance. Each stage's yield is reported separately.
  • I3. Two routes are built and compared side by side: an expensive one through a rich wordnet and a cheap one through plain dictionary glosses. The cheap one wins on coverage and the union wins overall.
  • I4. Every lexicon entry stores a confidence value and its source identifiers alongside the scores, which keeps the resource traceable.
  • I5. Coverage is measured separately from downstream F1, and both are compared against a majority baseline and a manually built competitor. This exposes when a precise but small resource loses on recall.

For Mova

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

  • I1 (world): Annotations attached to one WordNet release could be ported onto the release behind our ontology this way, with a POS check guarding against silent mismatches.
  • I2 (tokens): Aligning our lemmas with external lexicons (VerbNet, FrameNet, Wikidata labels) could use the same staged backoff, logging which stage produced each link so it stays explainable.
  • I3 (world): When compiling knowledge tables, we could build both a curated-link and a gloss-link version and keep their union, tagging each entry with its origin.
  • I4 (world): Compiled knowledge entries could carry a provenance record and a confidence value, so a reader can explain where a property of an entity came from.
  • I5 (learn): Lexicon or rule-set evaluations could always report coverage plus a trivial-baseline control, to show whether gains come from precision or from simply covering more input.

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

🎓New York University Abu DhabiAbu Dhabi · United Arab Emirates122 works · 4,602 citations🎓American University of BeirutBeirut · Lebanon34 works · 2,561 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksPsychology · 1,552 worksExperimental and Cognitive Psychology · 870 works

🧭 Topics 2Research subjects and languages the work is about.Sentiment & emotion · 972 worksSemitic languages · 2,908 works

Cite

Cite as Gilbert Badaro, Ramy Baly, Hazem Hajj, Nizar Habash and Wassim El-Hajj. 2014. A Large Scale Arabic Sentiment Lexicon for Arabic Opinion Mining. Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing (ANLP).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.3115/v1/w14-3623

@inproceedings{badaro2014large,
  title = {{A Large Scale Arabic Sentiment Lexicon for Arabic Opinion Mining}},
  author = {Gilbert Badaro, Ramy Baly, Hazem Hajj, Nizar Habash and Wassim El-Hajj},
  booktitle = {Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing (ANLP)},
  year = {2014},
  doi = {10.3115/v1/w14-3623},
  url = {https://doi.org/10.3115/v1/w14-3623}
}

References

23 references, 9 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: short papers-Volume 2. Association for Computational Linguistics.
  2. Abdul-Mageed, M., & Diab, M. (2012). Toward building a large-scale Arabic sentiment lexicon. In Proceedings of the 6th International Global WordNet Conference (pp. 18-22). find
  3. Abdul-Mageed, M., & Diab, M. (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), Reykjavik, Iceland Alhazmi, S., Black, W., & McNaught, J. (2013). Arabic SentiWordNet in Relation to SentiWordNet 3.0. 2180-1266, 4(1), 1-11.
  4. Baccianella, S., Esuli, A., & Sebastiani, F. (2010, May). SentiWordNet 3.0: An Enhanced Lexical Resource for Sentiment Analysis and Opinion Mining. In LREC (Vol. 10, pp. 2200-2204).
  5. Black, W., Elkateb, S., & Vossen, P. (2006). Introducing the Arabic wordnet project. In In Proceedings of the third International WordNet Conference (GWC-06). find
  6. Buckwalter, T. 2004. Buckwalter Arabic morphological analyzer version 2.0. LDC catalog number LDC2004L02, ISBN 1-58563-324-0 find
  7. Chen, Y., & Skienna, S. (2014). Building Sentiment Lexicons for All Major Languages. In Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Short Papers) (pp. 383-389). 2014 Association for Computational Linguistics. find
  8. Denecke, K. (2008, April). Using sentiwordnet for multilingual sentiment analysis. In Data Engineering Workshop, 2008. ICDEW 2008. IEEE 24th International Conference on (pp. 507-512). IEEE. find
  9. Diab, M., Al-Badrashiny, M., Aminian, M., Attia, M., Dasigi, P., Elfardy, H., Eskander, R., Habash, N., Hawwari, A., & Salloum, W. (2014). Tharwa: A Large Scale Dialectal Arabic - Standard Arabic - English Lexicon. In Proceedings of the Language Resources and Evaluation Conference (LREC), Reykjavik, Iceland.
  10. Elarnaoty, M., AbdelRahman, S., & Fahmy, A. (2012). A Machine Learning Approach for Opinion Holder Extraction in Arabic Language. International Journal of Artificial Intelligence & Applications, 3(2). ↗
  11. Esuli, A., & Sebastiani, F. (2005). Determining the semantic orientation of terms through gloss classification. In Proceedings of the 14th ACM international conference on Information and knowledge management (pp. 617-624). ACM. find
  12. Esuli, A., & Sebastiani, F. (2006). Sentiwordnet: A publicly available lexical resource for opinion mining. In Proceedings of LREC (Vol. 6, pp. 417-422).
  13. Graff, D., Maamouri, M., Bouziri, B., Krouna, S., Kulick, S., & Buckwalter, T. (2009). Standard Arabic Morphological Analyzer (SAMA) Version 3.1, 2009. Linguistic Data Consortium LDC2009E73. find
  14. Habash, N., & Rambow, O., (2005). Arabic tokenization, part-of-speech tagging and morphological disambiguation in one fell swoop. Proceedings of the 43rd Annual Meeting on Association for Computational Linguistics. Association for Computational Linguistics.
  15. Habash, N. (2010). Introduction to Arabic natural language processing. Synthesis Lectures on Human Language Technologies, 3(1), 1-187. ↗
  16. Liu, B. (2010). Sentiment Analysis and Subjectivity. Handbook of Natural Language Proccessing, 2:568. find
  17. Maamouri, M., Bies, A., Buckwalter, T., & Mekki, W. (2004, September). The penn arabic treebank: Building a large-scale annotated arabic corpus. In NEMLAR Conference on Arabic Language Resources and Tools (pp. 102-109). find
  18. Miller, G. A., Beckwith, R., Fellbaum, C., Gross, D., & Miller, K. J. (1990). Introduction to wordnet: An on-line lexical database. International journal of lexicography, 3(4), 235-244. ↗
  19. Ohana, B., & Tierney, B. (2009, October). Sentiment classification of reviews using SentiWordNet. In 9th. IT & T Conference (p. 13). find
  20. Pang, B., & Lee, L. (2004, July). A sentimental education: Sentiment analysis using subjectivity summarization based on minimum cuts. In Proceedings of the 42nd annual meeting on Association for Computational Linguistics (p. 271). Association for Computational Linguistics. ↗
  21. Pang, B., & Lee, L. (2005, June). Seeing stars: Exploiting class relationships for sentiment categorization with respect to rating scales. In Proceedings of the 43rd Annual Meeting on Association for Computational Linguistics (pp. 115-124). Association for Computational Linguistics. ↗
  22. Staiano, J., & Guerini, M. (2014). DepecheMood: a Lexicon for Emotion Analysis from Crowd-Annotated News. arXiv preprint arXiv:1405.1605. ↗
  23. Taboada, M., Brooke, J., Tofiloski, M., Voll, K., & Stede, M. (2011). Lexicon-based methods for sentiment analysis. Computational linguistics, 37(2), 267-307. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/badaro-2014-large-scale-arabic-sentiment-lexicon/