Elamur Research⌕
☰
home / papers / 2018 / Sentiment Analysis of Arabic Tweets: Feature Engineering and A Hybrid Approach

Sentiment Analysis of Arabic Tweets: Feature Engineering and A Hybrid Approach

Paper · 2018 · Nora Al-Twairesh, Hend Al-Khalifa, AbdulMalik Alsalman et al.

Summary

This work addresses sentiment classification of Arabic tweets. Social media Arabic is noisy, informal and heavily dialectal, and most Arabic NLP tools target MSA, so they handle it poorly. The authors therefore restrict themselves to one variant, the Saudi Dialect, and build a hybrid system on a large corpus of Saudi tweets. Lexicon knowledge enters a corpus-based classifier as features. These include presence of positive or negative words from three lexicons (the tweet-derived AraSenTi lexicon plus Arabic translations of MPQA and Bing Liu), negation and valence-shifter cues, emoticons and punctuation, and a TweetScore that sums AraSenTi intensity values. Morphological and POS features are deliberately left out. Feature groups are pruned by backward selection, which doubles as an ablation study. Best F1 is 69.9 for two-way, 61.63 for three-way and 55.07 for four-way classification. The AraSenTi features survive in every best model. On the Egyptian ASTD benchmark, the same features and selection procedure beat the published baseline. This shows that tweet-specific lexicons and careful feature pruning matter for dialectal sentiment analysis.

Key points

  • The three label schemes are two-way (pos/neg), three-way (adds neutral) and four-way (adds mixed), with best F1 scores of 69.9, 61.63 and 55.07.
  • Features from the Twitter-derived AraSenTi lexicon appear in the best feature set of every model. Emoticon and tweet-length features mostly hurt; removing the emoticon group gave a +23% gain in four-way classification.
  • POS and morphological features were excluded because taggers trained on MSA newswire had not helped earlier Arabic tweet studies. One prior study reported a 21% drop when using them.
  • Cross-dialect test (train Saudi, test Egyptian ASTD, four-way): F1 fell to 53.68.
  • When training on ASTD itself (89/11 split), the engineered features plus backward selection reached 65.8 F1. The n-gram benchmark of Nabil et al. scored 62.6.

Interesting ideas

  • I1. The method starts with every feature group, drops one group at a time and permanently removes any group whose absence improves the score, so model selection and ablation happen in one procedure.
  • I2. A lexicon harvested from the target genre itself beat lexicons translated from dictionaries, because it captures slang and non-standard spellings.
  • I3. Valence shifters get distinct roles: modal words tend to neutralize sentiment, and contrast words like "but" suggest two opposing evaluations around them.
  • I4. The authors commit to one language variant up front and then measure the cost of crossing variants (Saudi→Egyptian) as an explicit stress test.
  • I5. Lexicon-based knowledge (a summed intensity score) is fed as a feature into the learned model rather than run as a separate voter. Hybridization stays inside one model, and its contribution is still measurable.

For Mova

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

  • I1 (learn): The evaluation harness could ablate whole rule or feature families of a component and log which ones help or harm. This would give an explainable pruning record instead of a single opaque score.
  • I2 (world): Compiled knowledge such as verb classes or value words could be supplemented with entries mined from the actual text genres Mova reads, like fables or school tests, so coverage matches real usage.
  • I3 (frames): The prag crate could treat modals as stance dampeners and contrast connectives as markers of split evaluation across clauses, which helps with irony and implicature.
  • I4 (tree): When converting or training on new treebanks or registers, Mova should state which dialect each parser targets. It should also report cross-register scores as a standing check of how far the parser generalizes.
  • I5 (answer): Readers could take symbolic evidence, such as counts of matched frame cues, as explicit inputs to their scoring, so each answer can trace which knowledge source tipped it.

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, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 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 Nora Al-Twairesh, Hend Al-Khalifa, AbdulMalik Alsalman and Yousef Al-Ohali. 2018. Sentiment Analysis of Arabic Tweets: Feature Engineering and A Hybrid Approach. arXiv:1805.08533.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.1805.08533 · arXiv 1805.08533

@misc{altwairesh2018sentiment,
  title = {{Sentiment Analysis of Arabic Tweets: Feature Engineering and A Hybrid Approach}},
  author = {Nora Al-Twairesh, Hend Al-Khalifa, AbdulMalik Alsalman and Yousef Al-Ohali},
  year = {2018},
  doi = {10.48550/arxiv.1805.08533},
  eprint = {1805.08533},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1805.08533}
}

References

38 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Ahmed Abbasi, Hsinchun Chen, and Arab Salem. 2008. Sentiment analysis in multiple languages: Feature selection for opinion classification in Web forums. ACM Transactions on Information Systems (TOIS), 26(3):12. find
  2. Nawaf A. Abdulla, Nizar A. Ahmed, Mohammed A. Shehab, and Mahmoud Al-Ayyoub. 2013. Arabic sentiment analysis: Lexicon-based and corpus-based. In Applied Electrical Engineering and Computing Technologies (AEECT), pages 1–6. IEEE. find
  3. Muhammad Abdul-Mageed. 2015. Subjectivity and sentiment analysis of Arabic as a morophologically-rich language. Ph.D. thesis, INDIANA UNIVER-SITY. find
  4. Muhammad Abdul-Mageed, Mona Diab, and Sandra Kübler. 2014. SAMAR: Subjectivity and sentiment analysis for Arabic social media. Computer Speech & Language, 28(1):20–37, January. ↗
  5. Muhammad Abdul-Mageed, Mona T. Diab, and Mohammed 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, pages 587–591. Association for Computational Linguistics.
  6. Muhammad Abdul-Mageed, Sandra Kübler, and Mona Diab. 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, pages 19–28. Association for Computational Linguistics.
  7. Haifa K Aldayel and Aqil M Azmi. 2015. Arabic tweets sentiment analysis–a hybrid scheme. Journal of Information Science. in press. ↗
  8. Nora Al-Twairesh, Hend Al-Khalifa, and AbdulMalik Al-Salman. 2016. AraSenTi: Large-Scale Twitter-Specific Arabic Sentiment Lexicons. In Proceedings of the 54th Annual Meeting on Association for Computational Linguistics, Berlin, Germany. Association for Computational Linguistics.
  9. Nora Al-Twairesh, Hend Al-Khalifa, AbdulMalik Al-Salman, and Yousef Al-Ohali. 2017. AraSenTi-Tweet: A Corpus for Arabic Sentiment Analysis of Saudi Tweets. In Procedia Computer Science, volume 117, pages 63–72, Dubai. Elsevier. find
  10. Nora Al-Twairesh, Hend S. Al-Khalifa, and AbdulMalik Al-Salman. 2014. Subjectivity and Sentiment Analysis of Arabic : Trends and Challenges. In The 11th ACS/IEEE International Conference on Computer Systems and Applications (AICCSA’ 2014), Doha, Qatar. find
  11. Matthias Hagen Martin Potthast Michel Büchner and Benno Stein. 2015. Webis: An Ensemble for Twitter Sentiment Detection. In The 9th International Workshop on Semantic Evaluation, page 582, Denver, Colorado, USA, June. find
  12. Chih-Chung Chang and Chih-Jen Lin. 2011. LIBSVM: a library for support vector machines. ACM Transactions on Intelligent Systems and Technology (TIST), 2(3):27. find
  13. Kareem Darwish and Walid Magdy. 2014. Arabic Information Retrieval. Foundations and Trends in Information Retrieval, 7(4):239–342. ↗
  14. Eli Dresner and Susan C Herring. 2014. Emoticons and illocutionary force. In Perspectives on Theory of Controversies and the Ethics of Communication, pages 81–90. Springer. ↗
  15. Ali Farghaly and Khaled Shaalan. 2009. Arabic natural language processing: Challenges and solutions. ACM Transactions on Asian Language Information Processing (TALIP), 8(4):14. ↗
  16. Kevin Gimpel, Nathan Schneider, Brendan O’Connor, Dipanjan Das, Daniel Mills, Jacob Eisenstein, Michael Heilman, Dani Yogatama, Jeffrey Flanigan, and Noah A Smith. 2011. Part-of-speech tagging for twitter: Annotation, features, and experiments. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies: short papers-Volume 2, pages 42–47. Association for Computational Linguistics. ↗
  17. Alec Go, Richa Bhayani, and Lei Huang. 2009. Twitter sentiment classification using distant supervision. CS224N Project Report, Stanford:1–12. find
  18. Nizar Habash, Ramy Eskander, and Abdelati Hawwari. 2012. A morphological analyzer for Egyptian Arabic. In Proceedings of the Twelfth Meeting of the Special Interest Group on Computational Morphology and Phonology, pages 1–9. Association for Computational Linguistics.
  19. Nizar Y Habash. 2010. Introduction to Arabic natural language processing. Synthesis Lectures on Human Language Technologies, 3(1):1–187. ↗
  20. Mark Hall, Eibe Frank, Geoffrey Holmes, Bernhard Pfahringer, Peter Reutemann, and Ian H Witten. 2009. The WEKA data mining software: an update. ACM SIGKDD explorations newsletter, 11(1):10–18. ↗
  21. Trevor J. Hastie, Robert John Tibshirani, and Jerome H Friedman. 2011. The elements of statistical learning: data mining, inference, and prediction. Springer, Second edition. find
  22. Minqing Hu and Bing Liu. 2004. Mining and summarizing customer reviews. In Proceedings of the tenth ACM SIGKDD international conference on Knowledge discovery and data mining, pages 168– 177. ACM. find
  23. Svetlana Kiritchenko, Xiaodan Zhu, and Saif M Mohammad. 2014. Sentiment analysis of short informal texts. Journal of Artificial Intelligence Research, 50:723–762. ↗
  24. Mohammed Korayem, David Crandall, and Muhammad Abdul-Mageed. 2012. Subjectivity and sentiment analysis of arabic: A survey. In Advanced Machine Learning Technologies and Applications, pages 128– 139. Springer. find
  25. Saif M Mohammad, Svetlana Kiritchenko, and Xiaodan Zhu. 2013. NRC-Canada: Building the State-ofthe-Art in Sentiment Analysis of Tweets. In Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2013), page 321, Atlanta, Georgia, USA. find
  26. Ahmed Mourad and Kareem Darwish. 2013a. 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.WASSA 2013, pages 55–64.
  27. Ahmed Mourad and Kareem Darwish. 2013b. Subjectivity and Sentiment Analysis of Modern Standard Arabic and Arabic Microblogs. WASSA 2013:55.
  28. Hamdy Mubarak and Kareem Darwish. 2014. Using Twitter to collect a multi-dialectal corpus of Arabic. ANLP 2014:1.
  29. Mahmoud Nabil, Mohamed Aly, and Amir F Atiya. 2015. ASTD: Arabic Sentiment Tweets Dataset. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 2515–2519. ↗
  30. Alexander Pak and Patrick Paroubek. 2010. Twitter as a Corpus for Sentiment Analysis and Opinion Mining. In Proceedings of the Sixth International Conference on Language Resources and Evaluation (LREC 2010), Valleta,Malta. European Language Resources Association (ELRA).
  31. Arfath Pasha, Mohamed Al-Badrashiny, Ahmed El Kholy, Ramy Eskander, Mona Diab, Nizar Habash, Manoj Pooleery, Owen Rambow, and Ryan Roth. 2014. Madamira: A fast, comprehensive tool for morphological analysis and disambiguation of arabic. In Proceedings of the 9th International Conference on Language Resources and Evaluation, LREC 2014, Reykjavik, Iceland. European Language Resources Association (ELRA).
  32. Livia Polanyi and Annie Zaenen. 2006. Contextual valence shifters. In Computing attitude and affect in text: Theory and applications, pages 1–10. Springer. ↗
  33. Eshrag Refaee and Verena Rieser. 2014a. Subjectivity and Sentiment Analysis of Arabic Twitter Feeds with Limited Resources. In Workshop on Free/Open-Source Arabic Corpora and Corpora Processing Tools Workshop Programme, page 16. find
  34. Eshrag Refaee and Verena Rieser. 2014b. Subjectivity and Sentiment Analysis of Arabic Twitter Feeds with Limited Resources. In Workshop on Free/Open-Source Arabic Corpora and Corpora Processing Tools Workshop Programme, page 16. find
  35. Mohammad Salameh, Saif M Mohammad, and Svetlana Kiritchenko. 2015. Sentiment after translation: A case-study on arabic social media posts. In Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 767– 777.
  36. Maite Taboada, Julian Brooke, Milan Tofiloski, Kimberly Voll, and Manfred Stede. 2011. Lexicon-based methods for sentiment analysis. Computational linguistics, 37(2):267–307. ↗
  37. Theresa Wilson, Janyce Wiebe, and Paul Hoffmann. 2009. Recognizing contextual polarity: An exploration of features for phrase-level sentiment analysis. Computational linguistics, 35(3):399–433. ↗
  38. Xiaodan Zhu, Svetlana Kiritchenko, and Saif M Mohammad. 2014. Nrc-canada-2014: Recent improvements in the sentiment analysis of tweets. In Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 443–447. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/al-twairesh-2018-sentiment-analysis-arabic-tweets-feature/