Elamur Research⌕
☰
home / papers / 2014 / SANA: A Large Scale Multi-Genre, Multi-Dialect Lexicon for Arabic Subjectivity and Sentiment Analysis

SANA: A Large Scale Multi-Genre, Multi-Dialect Lexicon for Arabic Subjectivity and Sentiment Analysis

Paper · 2014 · Muhammad Abdul-Mageed and Mona Diab

Summary

Arabic sentiment research has lacked large polarity lexicons comparable to those available for English, and existing work has focused almost entirely on Modern Standard Arabic. The authors build SANA, a subjectivity and sentiment lexicon for Arabic. It spans several genres: newswire, chat, Twitter and YouTube. It also covers several varieties: MSA, Egyptian (EDA) and Levantine (LDA), with English glosses. Parts of it carry POS tags, diacritics, gender, number and rationality.

SANA was assembled in two ways:

  • Manual labeling: native speakers labeled two lexicons, SIFAAT (adjectives from the Penn Arabic Treebank) and HUDA (frequent words from Egyptian chat).
  • Automatic acquisition:
    • PMI-based association scores.
    • Machine translation of English lexicons: SentiWordNet 3, a YouTube lexicon and General Inquirer.
    • Pivoting English glosses through SAMA to collect several Arabic equivalents per entry.
    • Linking to the multi-dialect THARWA dictionary.

The result is a resource with many thousands of entries labeled Pos/Neg/Obj. The authors acknowledge that deduplication and quality control are incomplete and that SANA has not yet been fully tested in downstream sentiment systems. Its value lies in covering dialects and noisy social media text that MSA-only resources miss.

Key points

  • SIFAAT has 3,325 newswire adjectives from PATB parts 1–4 (617 Pos, 550 Neg, 2,158 Obj). Two annotators labeled them independently and then adjudicated disagreements.
  • HUDA has 4,905 high-frequency entries from about 11 million Yahoo Maktoob! Egypt chat turns collected in 2008 (1,900 Pos, 1,080 Neg, 1,925 Obj). A single annotator labeled them, and the text was not preprocessed.
  • From SWN3's 205,823 unique English entries, the Google Translation API returned Arabic output for 127,291 (61.84%). The YT lexicon (29,991 words from the 229,177-comment YuHC) was also translated.
  • ACT entries were matched to SAMA glosses (2,479 matches) and expanded to 6,965 Arabic entries. After 36 entries with conflicting labels were dropped, 6,929 remained. GI produced 4,269 Arabic entries (1,859 Pos, 2,410 Neg).
  • THARWA, a dictionary linking Egyptian, Levantine, MSA and English, was merged twice. One instance links through the translated SWN3; the other links through the manually labeled lexicons. This keeps automatically and manually derived material apart.

Interesting ideas

  • I1. Using English glosses as a pivot lets one labeled English word fan out to all target-language words sharing that gloss, and entries that end up with contradictory labels are simply discarded.
  • I2. The authors deliberately kept resources derived by hand separate from those derived automatically, building parallel instances rather than one blended lexicon, so quality can be judged by origin.
  • I3. Annotators were told to label prior polarity by a word's typical general use and its majority sense, not by personal experience or one particular context.
  • I4. For noisy chat and comments, the authors labeled the most frequent raw forms directly and squeezed letter repetitions above two down to two (cooool becomes cool), avoiding preprocessing that would itself make errors.
  • I5. PMI against self-assigned clues such as hashtags or emoticons can approximate polarity classes in social media without manual annotation.

For Mova

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

  • I1 (world): Polarity or class labels could be propagated from one WordNet/FrameNet sense to every lemma attached to it, with automatic removal of lemmas that receive conflicting values.
  • I2 (learn): Every silver annotation could carry a provenance tier (human, witness-parser agreement, LLM teacher), so evaluation and repair rules can weigh or filter by source.
  • I3 (frames): Out-of-context polarity priors built this way could serve as defaults, which irony and implicature detection then overrides when the context contradicts them.
  • I4 (tokens): A repeated-letter normalization step plus a frequency-ranked raw-form list could help the tokenizer and lemmatizer handle informal English without a fragile pipeline.
  • I5 (learn): Explicit surface cues could act as weak labels for inducing lexical priors, checked against negative controls before being trusted.

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

🎓George Washington UniversityWashington · United States51 works · 8,148 citations🎓Indiana University BloomingtonBloomington · United States37 works · 3,685 citations🎓Indiana UniversityBloomington · United States41 works · 3,663 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 Muhammad Abdul-Mageed and Mona Diab. 2014. SANA: A Large Scale Multi-Genre, Multi-Dialect Lexicon for Arabic Subjectivity and Sentiment Analysis. Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC'14).

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

@inproceedings{abdulmageed2014sana,
  title = {{SANA: A Large Scale Multi-Genre, Multi-Dialect Lexicon for Arabic Subjectivity and Sentiment Analysis}},
  author = {Muhammad Abdul-Mageed and Mona Diab},
  booktitle = {Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC'14)},
  year = {2014},
  doi = {10.63317/4xff9iiz2q7b},
  url = {https://doi.org/10.63317/4xff9iiz2q7b}
}

References

32 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Abbasi, A., Chen, H., and Salem, A. (2008). Sentiment analysis in multiple languages: Feature selection for opinion classification in web forums. ACM Trans. Inf. Syst., 26:1–34. find
  2. Abdul-Mageed, M. and Diab, M. (2011). Subjectivity and sentiment annotation of modern standard arabic newswire. In Proceedings of the 5th Linguistic Annotation Workshop, pages 110–118, Portland, Oregon, USA, June. Association for Computational Linguistics.
  3. Abdul-Mageed, M. and Diab, M. (2012a). AWATIF: A multi-genre corpus for modern standard arabic subjectivity and sentiment analysis. In Proceedings of LREC, volume 12.
  4. Abdul-Mageed, M. and Diab, M. (2012b). Toward building a large-scale arabic sentiment lexicon. In Proceedings of the 6th International Global WordNet Conference, Matsue, Japan, January. find
  5. Abdul-Mageed, M., Diab, M., and Korayem, M. (2011a). Subjectivity and sentiment analysis of modern standard arabic. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, pages 587–591, Portland, Oregon, USA, June. Association for Computational Linguistics.
  6. Abdul-Mageed, M., Korayem, M., and YoussefAgha, A. (2011b). ”yes we can?”: Subjectivity annotation and tagging for the health domain. In Proceedings of RANLP2011. find
  7. Abdul-Mageed, M., Kübler, S., and Diab, M. (2012). Samar: a system for subjectivity and sentiment analysis of arabic social media. In Proceedings of the 3rd Workshop in Computational Approaches to Subjectivity and Sentiment Analysis, WASSA ’12, pages 19–28, Stroudsburg, PA, USA. Association for Computational Linguistics.
  8. Abdul-Mageed, M., Diab, M., and Kbler, S. (2014). Samar: Subjectivity and sentiment analysis for arabic social media. Computer Speech Language, 28(1):20 – 37. ↗
  9. Baccianella, S., Esuli, A., and Sebastiani, F. (2010). Sentiwordnet 3.0: An enhanced lexical resource for sentiment analysis and opinion mining. In Seventh conference on International Language Resources and Evaluation, Malta. Retrieved May, volume 25, page 2010.
  10. Buckwalter, T. (2002). Arabic morphological analyzer version 2.0. find
  11. Linguistic Data Consortium, 2004. find
  12. Christopherson, K. M. (2007). The positive and negative implications of anonymity in internet social interactions:on the internet, nobody knows youre a dog. Computers in Human Behavior, 23(6):3038–3056. ↗
  13. Church, K. and Hanks, P. (1989). Word association norms, mutual information and lexicography. In Proceedings of the 27th Annual Conference of the ACL, pages 76–83. Association for Computational Linguistics. ↗
  14. Dave, K., Lawrence, S., and Pennock, D. (2003). Mining the peanut gallery: Opinion extraction and semantic classification of product reviews. In Proceedings of the 12th international conference on World Wide Web, pages 519–528. ACM. find
  15. Diab, M., Al-Badrashiny, M., Aminian, M., Attia, M., Dasigi, P., Elfardy, H., Eskander, R., Habash, N., Hawwari, A., Salloum, W., and Yusef, M. (2014). A large scale three-way egyptian arabic dictionary. find
  16. Esuli, A. and 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, pages 617– 624. ACM. find
  17. Fellbaum, C. (1998). WordNet: An electronic lexical database. The MIT press. find
  18. Graff, D., Maamouri, M., Bouziri, B., Krouna, S., Kulick, S., and Buckwalter, T. (2009). Standard Arabic Morphological Analyzer (SAMA) Version 3.1. Linguistic Data Consortium LDC2009E73. find
  19. Hatzivassiloglou, V. and McKeown, K. (1997). Predicting the semantic orientation of adjectives. In ACL-EACL, pages 174–181. find
  20. Heise, D. R. (2007). Expressive order: Confirming sentiments in social actions. Springer. ↗
  21. Herring, S. C. (2007). A faceted classification scheme for computer-mediated discourse. Language@ internet, 4(1):1–37. find
  22. Kamps, J., Marx, M., Mokken, R., and De Rijke, M. (2004). Using wordnet to measure semantic orientations of adjectives. find
  23. Kim, S. and Hovy, E. (2004). Determining the sentiment of opinions. In Proceedings of the 20th International Conference on Computational Linguistics, pages 1367–1373.
  24. Landis, J. and Koch, G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33(1):159. ↗
  25. Maamouri, M., Bies, A., Buckwalter, T., and Mekki, W. (2004). The penn arabic treebank: Building a large-scale annotated arabic corpus. In NEMLAR Conference on Arabic Language Resources and Tools, pages 102– 109. find
  26. Maamouri, M., Graff, D., Bouziri, B., Krouna, S., and Kulick, S. (2010). Ldc standard arabic morphological analyzer (sama) v. 3.1. LDC Catalog No. LDC2010L01. ISBN, pages 1–58563. find
  27. Miller, G. (1995). Wordnet: a lexical database for english. Communications of the ACM, 38(11):39–41.
  28. Mohammad, S. M. (2012). # emotional tweets. In Proceedings of the First Joint Conference on Lexical and Computational Semantics-Volume 1: Proceedings of the main conference and the shared task, and Volume 2: Proceedings of the Sixth International Workshop on Semantic Evaluation, pages 246–255. Association for Computational Linguistics. find
  29. Mullen, T. and Collier, N. (2004). Sentiment analysis using support vector machines with diverse information sources. In Proceedings of the Empirical Methods in Natural Language Processing (EMNLP) Conference, pages 412–418, Barcelona, Spain. find
  30. Stone, P. J., Dunphy, D. C., and Smith, M. S. (1966). The general inquirer: A computer approach to content analysis. find
  31. Turney, P. D. and Littman, M. L. (2003). Measuring praise and criticism: Inference of semantic orientation from association. ACM Transactions on Information Systems (TOIS), 21(4):315–346. ↗
  32. Turney, P. (2002). Thumbs up or thumbs down? semantic orientation applied to unsupervised classification of reviews. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics (ACL’02). ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/mageed-2014-sana-large-scale-multi-genre/