Elamur Research⌕
☰
home / papers / 2017 / Using objective words in the reviews to improve the colloquial arabic sentiment analysis

Using objective words in the reviews to improve the colloquial arabic sentiment analysis

Summary

Sentiment analysis of Arabic is hard because people write reviews in colloquial dialects that have no fixed spelling, morphology or vocabulary. Most work focuses on MSA. This paper looks at Jordanian colloquial reviews and asks whether "objective" words help with polarity classification. These are words with no explicit sentiment, which most models discard. The author builds two lexicons. The first holds colloquial sentiment words and multiword compound phrases. The second holds objective words, each given a sentiment-tendency value by an estimation method based on how the words appear in reviews. Features drawn from both lexicons are fed to an SVM, alone and combined with different n-gram models, to label each review as positive or negative. The reviews were gathered by hand from the JEERAN website. The combined approach beats two baseline models and reaches 95.6% accuracy. The work matters for two reasons: dialect resources for Arabic are scarce, and it questions the common practice, following Pang and Lee, of removing objective content before classifying sentiment.

Key points

  • Target: binary (positive/negative) polarity of Jordanian colloquial Arabic reviews, collected manually from JEERAN.
  • Two hand-built resources:
    • a lexicon of colloquial sentiment words plus compound phrases;
    • a lexicon of objective words, each with an estimated sentiment-tendency score.
  • Classifier: SVM (LIBSVM is cited). Lexicon-derived features are tested alone and together with n-gram models.
  • Best reported accuracy is 95.6%, higher than two baseline models.
  • Motivation comes from English work by Hung and Lin. They noted that over 90% of SentiWordNet entries are objective. Re-scoring those entries by their occurrence in positive and negative reviews improved accuracy by 4.10%.

Interesting ideas

  • I1. "Neutral" words are not necessarily neutral within a domain: a word's distribution across positive and negative reviews can reveal a hidden lean, and that lean can be measured instead of assumed to be zero.
  • I2. Treat multiword compound phrases as single sentiment units in the lexicon, rather than scoring their parts separately, because new colloquial compounds carry meanings their components do not.
  • I3. Hybrid design: a lexicon supplies interpretable features, and a learned classifier decides. This addresses the opposite biases of each approach: learners are conservative on positives full of factual statements, and lexicons over-react to a few positive words.
  • I4. Dialects keep generating new clitics, negation forms and stop-words, so a fixed stop-list or negation list goes stale and silently damages downstream analysis.

For Mova

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

  • I1 (world): Domain-conditioned evaluative leanings could be stored as revisable values on ontology entries (e.g., a product feature that tends to appear in complaints). Each value would stay traceable to the evidence that produced it.
  • I2 (frames): Evaluative idioms could be recognized as one unit over a UD subtree and given one attitude in the pragmatics layer. This avoids summing the polarity of each word.
  • I3 (learn): When evaluating any lexicon-driven component, compare it against both a pure-lexicon and a pure-learned baseline. This shows which bias the component actually fixes.
  • I4 (tokens): Negation and function-word inventories should be versioned, extensible data with coverage checks on new text, not hard-coded lists.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Topics Sentiment & emotion Semitic languages

 Institutions

Also on the byline: Centre de Recherche en Linguistique Appliquée.

Cite

Cite as Omar Al-Harbi. 2017. Using objective words in the reviews to improve the colloquial arabic sentiment analysis. arXiv:1709.08521 · International Journal on Natural Language Computing (IJNLC) Vol. 6, No.3, June 2017.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.5121/ijnlc · arXiv 1709.08521

@misc{alharbi2017using,
  title = {{Using objective words in the reviews to improve the colloquial arabic sentiment analysis}},
  author = {Omar Al-Harbi},
  year = {2017},
  doi = {10.5121/ijnlc},
  eprint = {1709.08521},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.5121/ijnlc}
}

References

44 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. H. Tang, S. Tan, and X. Cheng, “A survey on sentiment detection of reviews,” Expert Systems with Applications, vol. 36, no. 7, pp. 10 760–10 773, 2009. find
  2. M. Hu and B. Liu, “Mining and summarizing customer reviews,” in Proceedings of the tenth ACM SIGKDD international conference on Knowledge discovery and data mining. ACM, 2004, pp. 168– 177. find
  3. B. Liu, “Sentiment analysis and subjectivity.” Handbook of natural language processing, vol. 2, pp. 627–666, 2010. find
  4. W. Medhat, A. Hassan, and H. Korashy, “Sentiment analysis algorithms and applications: A survey,” Ain Shams Engineering Journal, vol. 5, no. 4, pp. 1093–1113, 2014. find
  5. B. Liu, “Sentiment analysis and opinion mining,” Synthesis lectures on human language technologies, vol. 5, no. 1, pp. 1–167, 2012. find
  6. M. Taboada, J. Brooke, M. To_loski, K. Voll, and M. Stede, “Lexicon-based methods for sentiment analysis,” Computational linguistics, vol. 37, no. 2, pp. 267–307, 2011. our card
  7. A. Kennedy and D. Inkpen, “Sentiment classi_cation of movie reviews using contextual valence shifters,” Computational intelligence, vol. 22, no. 2, pp. 110–125, 2006. find
  8. M. Yang, W. Tu, Z. Lu, W. Yin, and K.-P. Chow, “Lcct: a semisupervised model for sentiment classi_- cation,” in Human Language Technologies: The 2015 Annual Conference of the North American Chapter of the ACL. Association for Computational Linguistics (ACL)., 2015. find
  9. K. Shaalan, H. Bakr, and I. Ziedan, “Transferring egyptian colloquial dialect into modern standard arabic,” in International Conference on Recent Advances in Natural Language Processing (RANLP– 2007), Borovets, Bulgaria, 2007, pp. 525–529. find
  10. M. Abdul-Mageed, S. Kuebler, and M. Diab, “Samar: A system for subjectivity and sentiment analysis of social media arabic,” in Proceedings of the 3rdWorkshop on Computational Approaches to Subjectivity and Sentiment Analysis (WASSA), ICC Jeju, Republic of Korea, 2012. find
  11. R. Duwairi, R. Marji, N. Sha’ban, and S. Rushaidat, “Sentiment analysis in arabic tweets,” in Information and communication systems (icics), 2014 5th international conference on. IEEE, 2014, pp. 1–6. find
  12. H. ElSahar and S. R. El-Beltagy, “A fully automated approach for arabic slang lexicon extraction from microblogs,” in International Conference on Intelligent Text Processing and Computational Linguistics. Springer, 2014, pp. 79–91. find
  13. B. Pang and L. Lee, “A sentimental education: Sentiment analysis using subjectivity summarization based on minimum cuts,” in Proceedings of the 42nd annual meeting on Association for Computational Linguistics. Association for Computational Linguistics, 2004, p. 271. our card
  14. N. A. Abdulla, N. A. Ahmed, M. A. Shehab, and M. Al-Ayyoub, “Arabic sentiment analysis: Lexiconbased and corpus-based,” in Applied Electrical Engineering and Computing Technologies (AEECT), 2013 IEEE Jordan Conference on. IEEE, 2013, pp. 1–6. find
  15. R. M. Duwairi, “Sentiment analysis for dialectical arabic,” in Information and Communication Systems (ICICS), 2015 6th International Conference on. IEEE, 2015, pp. 166–170. find
  16. J. Stokes and A. Gorman, “Encyclopedia of the peoples of africa and the middle east, 2010,” Online Edition, The Safavid and Qajar dynasties, rulers in Iran from, vol. 1501, p. 707. find
  17. L. McLoughlin, Colloquial Arabic (Levantine). Routledge, 2009. find
  18. S. R. El-Beltagy and A. Ali, “Open issues in the sentiment analysis of arabic social media: A case study,” in Innovations in information technology (iit), 2013 9th international conference on. IEEE, 2013, pp. 215–220. find
  19. P. McCarthy and C. Boonthum-Denecke, Applied natural language processing. Information Science Reference, 2011. find
  20. M. A. M. E. Ahmed, “Alarge-scale computational processor of the arabic morphology, and applications,” Ph.D. dissertation, Faculty of Engineering, Cairo University Giza, Egypt, 2000. find
  21. N. Y. Habash, “Introduction to arabic natural language processing,” Synthesis Lectures on Human Language Technologies, vol. 3, no. 1, pp. 1–187, 2010. find
  22. R. Hetzron, The Semitic Languages. Routledge, 2013. find
  23. A. Farghaly and K. Shaalan, “Arabic natural language processing: Challenges and solutions,” ACM Transactions on Asian Language Information Processing (TALIP), vol. 8, no. 4, p. 14, 2009. find
  24. N. Habash and O. Rambow, “Magead: a morphological analyzer and generator for the arabic dialects,” in Proceedings of the 21st International Conference on Computational Linguistics and the 44th annual meeting of the Association for Computational Linguistics. Association for Computational inguistics, 2006, pp. 681–688. our card
  25. C. Miller, E. Al-Wer, D. Caubet, J. C. Watson et al., Arabic in the city: Issues in dialect contact and language variation. Routledge, 2007. find
  26. C. Hung and H.-K. Lin, “Using objective words in sentiwordnet to improve sentiment classi_cation for word of mouth,” IEEE Intelligent Systems, vol. 28, no. 2, pp. 47–54, 2013. find
  27. Y. Kaewpitakkun, K. Shirai, and M. Mohd, “Sentiment lexicon interpolation and polarity estimation of objective and out-of-vocabulary words to improve sentiment classi_cation on microblogging.” In PACLIC, 2014, pp. 204–213. find
  28. H. Amiri and T.-S. Chua, “Sentiment classi_cation using the meaning of words,” in Workshops at the Twenty-Sixth AAAI Conference on Arti_cial Intelligence, 2012. find
  29. K. Ghag and K. Shah, “Sentit_df–sentiment classi_cation using relative term frequency inverse document frequency,” International Journal of Advanced Computer Science & Applications, vol. 5, no. 2, 2014. find
  30. H. S. Ibrahim, S. M. Abdou, and M. Gheith, “Sentiment analysis for modern standard arabic and colloquial,” arXiv preprint arXiv:1505.03105, 2015. ↗
  31. A. Shoukry and A. Rafea, “A hybrid approach for sentiment classi_cation of egyptian dialect tweets,” in Arabic Computational Linguistics (ACLing), 2015 First International Conference on. IEEE, 2015, pp. 78–85. find
  32. S. R. El-Beltagy, “Niletmrg at semeval-2016 task 7: Deriving prior polarities for arabic sentiment terms,” Proceedings of SemEval, pp. 486–490, 2016. find
  33. F. E.-z. El-taher, A. A. Hammouda, and S. Abdel-Mageid, “Automation of understanding textual contents in social networks,” in Selected Topics in Mobile & Wireless Networking (MoWNeT), 2016 International Conference on. IEEE, 2016, pp. 1–7. find
  34. A. M. Azmi and S. M. Alzanin, “Aara’–a system for mining the polarity of saudi public opinion through e-newspaper comments,” Journal of Information Science, vol. 40, no. 3, pp. 398–410, 2014. find
  35. J. B. Salamah and A. Elkhli_, “Microblogging opinion mining approach for kuwaiti dialect,” Computing Technology and Information Management, vol. 1, no. 1, p. 9, 2016. find
  36. A. Mourad and K. Darwish, “Subjectivity and sentiment analysis of modern standard arabic and Arabic microblogs,” in Proceedings of the 4th workshop on computational approaches to subjectivity, sentiment and social media analysis, 2013, pp. 55–64. find
  37. Z. Zhang, X. Li, and Y. Chen, “Deciphering word-of-mouth in social media: Text-based metrics of consumer reviews,” ACM Transactions on Management Information Systems (TMIS), vol. 3, no. 1, p. 5, 2012. find
  38. B. Pang, L. Lee, and S. Vaithyanathan, “Thumbs up?: sentiment classi_cation using machine learning techniques,” in Proceedings of the ACL-02 conference on Empirical methods in natural language processing-Volume 10. Association for Computational Linguistics, 2002, pp. 79–86. find
  39. S.-M. Kim, P. Pantel, T. Chklovski, and M. Pennacchiotti, “Automatically assessing review helpfulness,” in Proceedings of the 2006 Conference on empirical methods in natural language processing. Association for Computational Linguistics, 2006, pp. 423–430. find
  40. A. Ghose and P. G. Ipeirotis, “Designing novel review ranking systems: predicting the usefulness and impact of reviews,” in Proceedings of the ninth international conference on Electronic commerce. ACM, 2007, pp. 303–310. find
  41. J. Liu, Y. Cao, C.-Y. Lin, Y. Huang, and M. Zhou, “Low-quality product review detection in opinion summarization.” in EMNLP-CoNLL, vol. 7, 2007, pp. 334–342. find
  42. P. Beineke, T. Hastie, and S. Vaithyanathan, “The sentimental factor: Improving review classi_cation via human-provided information,” in Proceedings of the 42nd annual meeting on association for computational linguistics. Association for Computational Linguistics, 2004, p. 263. find
  43. A. Agarwal, B. Xie, I. Vovsha, O. Rambow, and R. Passonneau, “Sentiment analysis of twitter data,” in Proceedings of the workshop on languages in social media. Association for Computational Linguistics, 2011, pp. 30–38. our card
  44. C.-C. Chang and C.-J. Lin, “Libsvm: a library for support vector machines,” ACM Transactions on Intelligent Systems and Technology (TIST), vol. 2, no. 3, p. 27, 2011. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/al-harbi-2017-using-objective-words-reviews-improve/