Elamur Research⌕
☰
home / papers / 2018 / ARB-SEN at SemEval-2018 Task1: A New Set of Features for Enhancing the Sentiment Intensity Prediction in Arabic Tweets

ARB-SEN at SemEval-2018 Task1: A New Set of Features for Enhancing the Sentiment Intensity Prediction in Arabic Tweets

Paper · 2018 · El Moatez Billah Nagoudi

Summary

The paper tackles valence intensity regression for Arabic tweets (the V-reg Arabic sub-task of SemEval-2018 Task1: Affect in Tweets). The goal is to predict a real-valued sentiment strength for each tweet rather than just a polarity label. ARB-SEN cleans tweets by stripping user mentions, RTs, URLs and diacritics, splitting hashtags into words, and normalising interchangeable Arabic letters. It then builds four groups of hand-crafted features: sentiment lexicon scores, a negation flag, word-embedding-derived lexicon scores, and emoticon/emoji indicators. These features feed two scikit-learn regressors, Linear Regression and SVR, trained on the organisers' training and development tweets. On the 731-tweet test set, SVR with all features reaches a Pearson correlation of 0.720, placing 6th. LR with all features reaches 0.617. An incremental ablation shows that every feature group helps, with the embedding and emoji features giving the largest gains. The work shows that a small, transparent feature set can be competitive in a low-data, morphologically rich setting.

Key points

  • Four lexicons supply the SLF features:
    • the SemEval-2016 Arabic lexicon of Kiritchenko et al. (1,168 words plus 198 phrases);
    • the Mohammad & Kiritchenko co-occurrence lexicon (~43k entries);
    • ArabSenti (14k manually labelled words);
    • a dialectal lexicon of 480+ words (Refaee & Rieser).
  • Each tweet gets sum, average, min and max lexicon scores. Words absent from a lexicon are skipped.
  • The WEF features rely on embeddings from Zahran et al.'s (2015) Arabic CBOW model. Each word is scored by the averaged lexicon scores of its 5 nearest neighbours.
  • The negation feature (NF) is a binary flag based on five MSA negation words.
  • Training data: 933 training and 139 development tweets.
  • Cumulative ablation, Pearson correlation:
FeaturesLRSVR
SLF0.5230.647
+NF0.5240.649
+WEF0.5610.683
+EEF0.6170.720

The baseline scores −0.052.

Interesting ideas

  • I1. Back off from a sparse lexicon through embedding neighbours: score each word by averaging the known scores of its five closest vectors, which extends coverage to words the lexicon lacks.
  • I2. Summarise per-word scores with sum, mean, min and max together, so a single strongly charged word is not washed out by the average.
  • I3. A tweet-level negation flag added almost nothing (+0.001/+0.002), which suggests negation without scope is nearly useless as a signal.
  • I4. Report ablations cumulatively, adding one feature group at a time for both a weak and a strong learner, so a feature's contribution can be checked for consistency across models.
  • I5. Break hashtags into their component words before analysis, so the sentiment-bearing content inside a tag becomes visible to the downstream features.

For Mova

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

  • I1 (world): the compiled value/valence knowledge could use a neighbour-based fallback for out-of-lexicon lemmas, storing the neighbour list as the explanation and flagging the value as derived, not attested.
  • I2 (frames): keeping both the extreme positive and extreme negative item in a clause would give the irony detector a cheap signal for sharp polarity contrast inside one utterance.
  • I3 (frames): attach negation to the specific predicate it governs in the UD tree, and flip the polarity only of that frame, rather than marking the whole text.
  • I4 (learn): evaluation runs could add each repair rule or knowledge source stepwise under two parsers, confirming that a gain is not tied to one model.
  • I5 (tokens): the en tokenizer could segment hashtags and camel-cased handles into words before tagging and lemmatisation, keeping the original span for traceability.

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

🎓University of ElouedEl Oued · Algeria2 works · 6 citations

 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 El Moatez Billah Nagoudi. 2018. ARB-SEN at SemEval-2018 Task1: A New Set of Features for Enhancing the Sentiment Intensity Prediction in Arabic Tweets. Proceedings of the 12th International Workshop on Semantic Evaluation.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/s18-1055

@inproceedings{nagoudi2018arb,
  title = {{ARB-SEN at SemEval-2018 Task1: A New Set of Features for Enhancing the Sentiment Intensity Prediction in Arabic Tweets}},
  author = {El Moatez Billah Nagoudi},
  booktitle = {Proceedings of the 12th International Workshop on Semantic Evaluation},
  year = {2018},
  doi = {10.18653/v1/s18-1055},
  url = {https://doi.org/10.18653/v1/s18-1055}
}

References

26 references, 15 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. 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.
  2. Nizar Y Habash. 2010. Introduction to arabic natural language processing. Synthesis Lectures on Human Language Technologies, 3(1):1–187. ↗
  3. Svetlana Kiritchenko, Saif M. Mohammad, and Mohammad Salameh. 2016. Semeval-2016 task 7: Determining sentiment intensity of english and arabic phrases. In Proceedings of the International Workshop on Semantic Evaluation, SemEval ’16, San Diego, California, June.
  4. Muhammad Abdul-Mageed, Mona Diab, and Sandra Kübler. 2014. Samar: Subjectivity and sentiment Jonathon Read. 2005. Using emoticons to reduce dependency in machine learning techniques for sentiment classification. In Proceedings of the ACL student research workshop, pages 43–48. Association for Computational Linguistics. find
  5. Nele Lenze. 2017. Social media in the arab world: Communication and public opinion in the gulf states. European Journal of Communication, 32(1):77–79. find
  6. Tomas Mikolov, Wen-tau Yih, and Geoffrey Zweig. 2013. Linguistic regularities in continuous space word representations. In Hlt-naacl, volume 13, pages 746–751.
  7. Eshrag Refaee and Verena Rieser. 2014. An arabic twitter corpus for subjectivity and sentiment analysis. In LREC, pages 2268–2273.
  8. Sara Rosenthal, Alan Ritter, Preslav Nakov, and Veselin Stoyanov. 2014. Semeval-2014 task 9: Sentiment analysis in twitter. In Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 73–80, Dublin, Ireland, August. Association for Computational Linguistics and Dublin City University.
  9. Saif M Mohammad and Felipe Bravo-Marquez. 2017. Wassa-2017 shared task on emotion intensity. arXiv preprint arXiv:1708.03700. ↗
  10. Saif M Mohammad, Parinaz Sobhani, and Svetlana Kiritchenko. 2017. Stance and sentiment in tweets. ACM Transactions on Internet Technology (TOIT), 17(3):26. ↗
  11. Saif M Mohammad, Felipe Bravo-Marquez, Mohammad Salameh, and Svetlana Kiritchenko. 2018a. Semeval-2018 task 1: Affect in tweets. In Proceedings of International Workshop on Semantic Evaluation (SemEval-2018), New Orleans, LA, USA.
  12. Sara Rosenthal, Preslav Nakov, Svetlana Kiritchenko, Saif Mohammad, Alan Ritter, and Veselin Stoyanov. 2015. Semeval-2015 task 10: Sentiment analysis in twitter. In Proceedings of the 9th international workshop on semantic evaluation (SemEval 2015), pages 451–463.
  13. Saif M. Mohammad, Felipe Bravo-Marquez, Mohammad Salameh, and Svetlana Kiritchenko. 2018b. Semeval-2018 Task 1: Affect in tweets. In Proceedings of International Workshop on Semantic Evaluation (SemEval-2018), New Orleans, LA, USA.
  14. Sara Rosenthal, Noura Farra, and Preslav Nakov. 2017. SemEval-2017 task 4: Sentiment analysis in Twitter. In Proceedings of the 11th International Workshop on Semantic Evaluation, SemEval ’17, Vancouver, Canada, August. Association for Computational Linguistics.
  15. Saif M Mohammad. 2016. Sentiment analysis: Detecting valence, emotions, and other affectual states from text. In Emotion measurement, pages 201–237. Elsevier. find
  16. Mohammad Salameh Saif M. Mohammad and Svetlana Kiritchenko. 2016. Sentiment lexicons for arabic social media. In Proceedings of 10th edition of the the Language Resources and Evaluation Conference (LREC), Portorož, Slovenia.
  17. Ahmed Mourad and Kareem Darwish. 2013. 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, pages 55–64.
  18. Mohammad Salameh, Saif 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.
  19. El Moatez Billah Nagoudi, Jérémy Ferrero, Didier Schwab, and Hadda Cherroun. 2017. Word embedding-based approaches for measuring semantic similarity of arabic-english sentences. In The 6th International Conference on Arabic Language Processing, pages 19–33. Springer. find
  20. F. Salem. 2017. The arab social media report 2017: Social media and the internet of things: Towards data-driven policymaking in the arab world. Vol. 7. find
  21. Feixiang Wang, Zhihua Zhang, and Man Lan. 2016. Ecnu at semeval-2016 task 7: An enhanced supervised learning method for lexicon sentiment intensity ranking. In Proceedings of the 10th International Workshop on Semantic Evaluation (SemEval-2016), pages 491–496. find
  22. Preslav Nakov, Zornitsa Kozareva, Alan Ritter, Sara Rosenthal, Veselin Stoyanov, and Theresa Wilson. 2013. Semeval-2013 task 2: Sentiment analysis in twitter.
  23. Wiesław Wolny. 2016. Sentiment analysis of twitter data using emoticons and emoji ideograms. Studia Ekonomiczne, 296:163–171. find
  24. Preslav Nakov, Alan Ritter, Sara Rosenthal, Fabrizio Sebastiani, and Veselin Stoyanov. 2016. Semeval-2016 task 4: Sentiment analysis in twitter. In Proceedings of the 10th International Workshop on Semantic Evaluation (SemEval-2016), pages 1–18.
  25. Mohamed A Zahran, Ahmed Magooda, Ashraf Y Mahgoub, Hazem Raafat, Mohsen Rashwan, and Amir Atyia. 2015. Word representations in vector space and their applications for arabic. In International Conference on Intelligent Text Processing and Computational Linguistics, pages 430–443. Springer. find
  26. Saad Ranginwala and Alexander J Towbin. 2017. The power of promotion: using social media to promote a radiology department. Academic Radiology, 24(4):488–496. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/nagoudi-2018-arb-sen-semeval-2018-task1/