Elamur Research⌕
☰
home / papers / 2013 / Exploring the Effects of Word Roots for Arabic Sentiment Analysis

Exploring the Effects of Word Roots for Arabic Sentiment Analysis

Paper · 2013 · Shereen Oraby, Yasser El-Sonbaty and Mohamad Abou El-Nasr

Summary

The paper asks whether reducing Arabic words to their consonantal roots helps sentiment classification. Arabic builds many related words from one root plus patterns and affixes, so a root seen in positive or negative contexts could cover many unseen derived forms. The authors add root features to plain unigram features in a basic two-class polarity classifier. They compare three rooting tools: the Khoja stemmer, the ISRI stemmer and Tashaphyne. They test on two very different corpora. The first is the newswire PATB Part 1 subset with sentiment annotation from Abdul-Mageed et al. (2011). The second is the OCA movie-review corpus from Rushdi-Saleh et al. (2011), which contains slang and dialect. Root features improved results on both datasets.

  • On PATB, Khoja roots gave the largest gain and also beat the earlier published F1.
  • On OCA, Tashaphyne gave a small gain over the earlier accuracy figure.

The work matters because it shows a cheap, language-specific normalization step can help in morphologically rich languages without domain-specific lexicons.

Key points

  • Features and tools: only unigrams, and unigrams combined with roots, were evaluated. The roots came from Khoja, ISRI and Tashaphyne.
  • PATB (newswire, MSA):
    • Khoja raised F1 from 56.3 to 61.2 and accuracy from 63.8 to 68.5.
    • This exceeds the earlier F1 baseline of 57.8.
    • All three rooters beat plain unigrams.
  • OCA (movie reviews):
    • Tashaphyne reached 93.2 accuracy and 93.4 F1, against 92.6 for plain unigrams and an earlier baseline of 90.0.
    • Khoja slightly lowered accuracy (92.2).
  • Which tool won: the best rooter differed by corpus. The authors link this to formality, dialect use and how many words can be rooted. The two winners both check candidate roots against a root dictionary.
  • Acknowledged limits: root matching alone cannot handle negation or intensifiers. The usual English trick of negating everything up to the next punctuation mark fits Arabic poorly because of its freer word order.

Interesting ideas

  • I1. Keep the surface word as a feature and add the root next to it, so the model can fall back on the shared derivational family when a specific form is rare or unseen.
  • I2. Validate affix-stripping output against a dictionary of known roots; the dictionary-backed stemmers performed best here.
  • I3. Run several normalization tools under one fixed classifier to isolate how much the tool choice matters, and expect the winner to change with domain.
  • I4. Test the same method on a formal corpus and an informal, dialect-heavy corpus to check that a gain is general rather than domain-bound.
  • I5. The paper points out that word-level polarity breaks under negation, and that linear punctuation-bounded scope fails when word order is free.

For Mova

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

  • I1 (tokens): alongside each lemma, the en crate could expose a derivational-family key (e.g. succeed/success/successful via WordNet derivational links), letting later stages share evidence across related words.
  • I2 (tokens): the morphology analyzer could accept a candidate stem or lemma only if the lexicon knows it, and otherwise keep the surface form, so lemmas stay explainable.
  • I3 (learn): competing lemmatizers or analyzers could be scored as separate witnesses per register, instead of assuming one tool is best everywhere.
  • I4 (learn): evaluation suites could pair formal text (school tests) with informal text (dialogue in tales) so improvements are confirmed on both registers.
  • I5 (frames): prag could compute negation and intensifier scope from the UD tree (the neg/advmod attachment) rather than from a token window, which keeps polarity on the correct frame.

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

🎓Arab Academy for Science, Technology, and Maritime TransportAlexandria · Egypt5 works · 50 citations

 Fields, topics and concepts

📚 Shelves 4The 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 works

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

Cite

Cite as Shereen Oraby, Yasser El-Sonbaty and Mohamad Abou El-Nasr. 2013. Exploring the Effects of Word Roots for Arabic Sentiment Analysis. Proceedings of the Sixth International Joint Conference on Natural Language Processing.

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

@inproceedings{oraby2013exploring,
  title = {{Exploring the Effects of Word Roots for Arabic Sentiment Analysis}},
  author = {Shereen Oraby, Yasser El-Sonbaty and Mohamad Abou El-Nasr},
  booktitle = {Proceedings of the Sixth International Joint Conference on Natural Language Processing},
  year = {2013},
  url = {https://aclanthology.org/I13-1054/}
}

References

30 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. A. Abbasi, H. Chen, and A. Salem. 2008. Sentiment analysis in multiple languages: feature selection for opinion classification in web forums. ACM Trans. Inf. Syst., 26(3):12:1–12:34, June. find
  2. M. Abdul-Mageed and M. Diab. 2011. Subjectivity and sentiment annotation of modern standard arabic newswire. In Proceedings of the 5th Linguistic Annotation Workshop, LAW V ’11, pages 110– 118, Stroudsburg, PA, USA. Association for Computational Linguistics.
  3. M. Abdul-Mageed, M. Diab, and M. Korayem. 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, HLT ’11, pages 587–591, Stroudsburg, PA, USA. Association for Computational Linguistics.
  4. L. Albraheem and H. Al-Khalifa. 2012. Exploring the problems of sentiment analysis in informal arabic. In Proceedings of the 14th International Conference on Information Integration and Web-based Applications; Services, IIWAS ’12, pages 415–418, New York, NY, USA. ACM. find
  5. R. Bruce and J. Wiebe. 1999. Recognizing subjectivity: a case study in manual tagging. Nat. Lang. Eng., 5(2):187–205, June. ↗
  6. Q. Bsoul and M. Mohd. 2011. Effect of isri stemming on similarity measure for arabic document clustering. In Information Retrieval Technology, volume 7097 of Lecture Notes in Computer Science, pages 584–593. Springer Berlin Heidelberg. find
  7. E. Cambria, B. Schuller, Y. Xia, and C. Havasi. 2013. New avenues in opinion mining and sentiment analysis. Intelligent Systems, IEEE, 28(2):15–21. ↗
  8. K. Dave, S. Lawrence, and D. Pennock. 2003. Mining the peanut gallery: opinion extraction and semantic classification of product reviews. In Proceedings of the 12th international conference on World Wide Web, WWW ’03, pages 519–528, New York, NY, USA. ACM. find
  9. M. Grassi, E. Cambria, A. Hussain, and F. Piazza. 2011. Sentic web: A new paradigm for managing social media affective information. Cognitive Computation, 3(3):480–489. ↗
  10. N. Habash. 2008. Introduction to Arabic Natural Language Processing. Synthesis lectures on human language technologies. Morgan & Claypool Publishers. find
  11. T. Joachims. 2008. Svmlight: Support vector machine, http://svmlight.joachims.org/. ↗
  12. S. Khoja and R. Garside. 1999. Stemming arabic text (tech. rep.). Computer Department, Lancaster University, Lancaster. find
  13. S. Kim and E. Hovy. 2004. Determining the sentiment of opinions. In Proceedings of the 20th international conference on Computational Linguistics, COLING ’04, Stroudsburg, PA, USA. Association for Computational Linguistics.
  14. L. Larkey and M. Connell. 2001. Arabic information retrieval at umass in trec-10. University of Massachusetts. find
  15. B. Liu. 2012. Sentiment Analysis and Opinion Mining. Morgan & Claypool Publishers. ↗
  16. M. Maamouri, A. Bies, and W. Mekki. 2004. The penn arabic treebank: building a large-scale annotated arabic corpus. pages 102–109. find
  17. M. Maamouri, A. Bies, S. Kulick, F. Gaddeche, W. Mekki, S. Krouna, B. Bouziri, and W. Zaghouani. 2010. Arabic treebank: Part 1 v 4.1. ↗
  18. A. Magdy, M. Kholief, and Y. El-Sonbaty. 2013. Qasim: Arabic language question answer selection in machines. Conference and Labs of the Evaluation Forum. find
  19. D. Olsher. 2012. Full spectrum opinion mining: integrating domain, syntactic and lexical knowledge. In Data Mining Workshops (ICDMW), 2012 IEEE 12th International Conference on, pages 693–700. find
  20. B. Pang, L. Lee, and S. Vaithyanathan. 2002. Thumbs up?: sentiment classification using machine learning techniques. In Proceedings of the ACL-02 conference on empirical methods in natural language processing - Volume 10, EMNLP ’02, pages 79–86, Stroudsburg, PA, USA. Association for Computational Linguistics. ↗
  21. M. Rushdi-Saleh, M. Martin-Valdivia, L. Urena-Lopez, and J. Perea-Ortega. 2011. Oca: Opinion corpus for arabic. J. Am. Soc. Inf. Sci. Technol., 62(10):2045– 2054, October. ↗
  22. K. Ryding. 2006. A Reference Grammar of Modern Standard Arabic. Reference Grammars. Cambridge University Press, New York. ↗
  23. D. Sanjiv and M. Chen. 2001. Yahoo! for amazon: extraction market sentiment from stock message boards. In Proceedings of the 8th Asia Pacific Finance Association Annual Conference. find
  24. A. Shoukry and A. Rafea. 2012. Sentence-level arabic sentiment analysis. In Collaboration Technologies and Systems (CTS), 2012 International Conference on, pages 546–550. find
  25. StudyQuran. 2004. Project root list online. 2004. http://www.studyquran.co.uk/prlonline.htm. ↗
  26. K. Taghva, R. Elkhoury, and J. Coombs. 2005. Arabic stemming without a root dictionary. In Proceedings of the International Conference on Information Technology: Coding and Computing (ITCC’05) - Volume I - Volume 01, ITCC ’05, pages 152–157, Washington, DC, USA. IEEE Computer Society. find
  27. Tashaphyne. 2010. Arabic light stemmer, 0.2. 2010. http://tashaphyne.sourceforge.net/. ↗
  28. P. Turney and M. Littman. 2003. Measuring praise and criticism: Inference of semantic orientation from association. ACM Trans. Inf. Syst., 21(4):315–346, October. ↗
  29. J. Wiebe, R. Bruce, and T. O’Hara. 1999. Development and use of a gold-standard data set for subjectivity classifications. In Proceedings of the 37th annual meeting of the Association for Computational Linguistics on Computational Linguistics, ACL ’99, pages 246–253, Stroudsburg, PA, USA. Association for Computational Linguistics.
  30. H. Yu and V. Hatzivassiloglou. 2003. Towards answering opinion questions: separating facts from opinions and identifying the polarity of opinion sentences. In Proceedings of the 2003 conference on Empirical methods in natural language processing, EMNLP ’03, pages 129–136, Stroudsburg, PA, USA. Association for Computational Linguistics.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/oraby-2013-exploring-effects-word-roots-arabic/