Elamur Research⌕
☰
home / papers / 2020 / LEBANONUPRISING: a thorough study of Lebanese tweets

LEBANONUPRISING: a thorough study of Lebanese tweets

Paper · 2020 · Reda Khalaf and Mireille Makary

Summary

The paper presents what the authors describe as the first sentiment study of tweets in spoken Lebanese Arabic, a dialect with no prior labelled resources. Its data comes from the LebanonUprising hashtag that trended during the October 2019 socio-economic protests. The authors filtered a public Kaggle collection down to Lebanese-dialect tweets and had them hand-labelled with a custom web tool. To normalise dialect spellings, they built a Lebanese–MSA mapping dictionary. Tweets were then cleaned, stripped of stopwords and stemmed. Five scikit-learn classifiers were trained for binary positive/negative prediction and applied to a second set of tweets collected months later. All classifiers performed at a very similar level. The authors also tested whether particular emojis can signal sarcastic or funny tweets and report satisfactory results. A user-overlap analysis showed that the hashtag's audience largely changed over time. The work matters as a starting resource for an under-served Arabic dialect, though its bag-of-words pipeline is basic.

Key points

  • TDS (training set): 21,529 Lebanese tweets, filtered from about 100K tweets collected by Abed Khooli on 18–21 October 2019 and shared on Kaggle. PDS (prediction set): 24,798 tweets collected from May to 8 August 2020 with workbench data.
  • Annotation: five annotators used six labels (sarcastic, angry, negative, neutral, funny, positive). These were collapsed into a binary scheme with neutral tweets dropped, giving 8,943 negative and 12,586 positive tweets.
  • Pipeline: Lebanese→MSA dictionary mapping, then removal of links, emojis, punctuation and stopwords, then the Snowball stemmer.
  • Classifiers: MultinomialNB, SVM, KNN, SGDClassifier and Logistic Regression, tuned with grid search on an 85/15 split. Their accuracies were close to one another.
  • User overlap: TDS had 38,322 users with retweets and 9,316 without. PDS had 20,850 and 5,406. Only 3,425 users appeared in both, about 5%.

Interesting ideas

  • I1. Map dialect spelling variants onto one standard form before stemming, so that variant spellings of the same word count as one feature.
  • I2. Annotate with fine-grained labels, collapse them into a coarse task for training, and keep flags for the subtypes (sarcastic, funny) to test later hypotheses.
  • I3. Use the presence of specific emojis as a simple, transparent cue for sarcasm or humour.
  • I4. Train on one time period and predict on a later one to see how a model behaves under topic and audience drift.

For Mova

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

  • I1 (tokens): A hand-curated map from nonstandard or informal English spellings to canonical forms could run before the lemmatizer, and each rewrite could be logged so it stays explainable.
  • I2 (learn): Silver or teacher annotations could store detailed tags even when training targets a simpler scheme, so later experiments need no relabelling.
  • I3 (frames): The irony module could treat such surface markers as explicit, explainable evidence that feeds an irony or implicature decision rather than deciding it alone.
  • I4 (learn): Evaluation could include held-out text from a different period or source as a shift test alongside in-domain scores.

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

Also on the byline: Lebanese International University · Lebanese International University.

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 9,086 worksArtificial Intelligence · 8,889 worksPsychology · 798 worksExperimental and Cognitive Psychology · 463 works

🧭 Topics 1Research subjects and languages the work is about.Semitic languages · 1,484 works

Cite

Cite as Reda Khalaf and Mireille Makary. 2020. LEBANONUPRISING: a thorough study of Lebanese tweets. arXiv:2009.14459.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2009.14459 · arXiv 2009.14459

@misc{khalaf2020lebanonuprising,
  title = {{LEBANONUPRISING: a thorough study of Lebanese tweets}},
  author = {Reda Khalaf and Mireille Makary},
  year = {2020},
  doi = {10.48550/arxiv.2009.14459},
  eprint = {2009.14459},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2009.14459}
}

References

16 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Kouloumpis, Efthymios., Wilson, T., Moore, J. (2011) Twitter sentiment analysis: the good the bad and the OMG! ICWSM 11, 538–541 find
  2. Ibrahim, H.S., Abdou, S.M., Gheith, M. (2015) Sentiment analysis for modern standard Arabic and colloquial. arXiv preprint, arXiv:1505.03105 ↗
  3. Haifa, Kfir & Azmi, Aqil. (2015). Arabic tweets sentiment analysis - A hybrid scheme. Journal of Information Science. 42. 10.1177/0165551515610513. ↗
  4. Liu, Bing (2010) Sentiment Analysis and Subjectivity. Handbook of Natural Language Processing, Second Edition, (editors: N. Indurkhya and F. J. Damerau) find
  5. Wiesław, Wolny (2016) Emotion analysis of twitter data that use emoticons and emoji ideogram 25th International Conference On Information Systems Development (ISD2016 Katowice) find
  6. Parrott, W (2001). Emotions in social psychology: Essential readings. Psychology Press. Excluding retweets 9,316 5,406 find
  7. Korayem, M., Crandall, D., Abdul-Mageed, M (2012) Subjectivity and sentiment analysis of Arabic: a survey. In: Hassanien, A.E., Salem, A.-B.M., Ramadan, R., Kim, T. (eds.) AMLTA 2012. CCIS, vol. 322, pp. 128–139. Springer, Heidelberg. doi:10.1007/978-3-642-35326-0 14 ↗
  8. Alwakid, Ghadah, Osman, Taha, Hughes-Roberts, Thomas (2017) Challenges in Sentiment Analysis for Arabic Social Networks 3rd International Conference on Arabic Computational Linguistics, ACLing 2017, Dubai, UAE find
  9. Duwairi, Rehab & Marji, Raed & Sha'ban, Narmeen & Rushaidat, Sally. (2014). Sentiment Analysis in Arabic tweets. 2014 5th International Conference on Information and Communication Systems, ICICS 2014. 1-6. 10.1109/IACS.2014.6841964. ↗
  10. Boser, Bernhard E., Guyon Isabelle M., & Vapnik, Vladimir N. (1992). A training algorithm for optimal margin classifiers. In Proceedings of the fifth annual workshop on Computational learning theory (COLT '92). Association for Computing Machinery, New York, NY, USA, 144– 152. DOI:https://doi.org/10.1145/130385.130401 ↗
  11. Lewis, David. (1998) Naive Bayes at forty: the independence assumption in information retrieval. In Machine Learning: ECML-98, Proceedings of the 10th European Conference on Machine Learning, Chemnitz, Germany (pp. 4–15). Berlin: Springer. find
  12. Fix, Evelyn, and J. L. Hodges. (1989) Discriminatory Analysis. Nonparametric Discrimination: Consistency Properties. International Statistical Review / Revue Internationale De Statistique 57, no. 3: 238-47. Accessed August 26, 2020. doi:10.2307/1403797. ↗
  13. Abdullah, Malak & Hadzikadic, Mirsad. (2017). Sentiment Analysis on Arabic Tweets: Challenges to Dissecting the Language. 191-202. 10.1007/978-3-319-58562-8_15. ↗
  14. Al-Twairesh, Nora, Hend Al-Khalifa, AbdulMalik Alsalman, and Yousef Al-Ohali. (2018) Sentiment analysis of arabic tweets: Feature engineering and a hybrid approach. arXiv preprint arXiv:1805.08533. ↗
  15. Ruder, Sebastian. (2016). An overview of gradient descent optimization algorithms. ArXiv Preprint ArXiv:1609.04747. ↗
  16. Subramanian, Jayashree & Sridharan, Varun & Shu, Kai & Liu, Huan. (2019). Exploiting Emojis for Sarcasm Detection. Authors Reda Khalaf: Master student in the Computer Science and Information Technology Department at the Lebanese International University Mireille Makary, PhD: Lecturer in the Computer Science and Information Technology Department at the Lebanese International University with research focus on machine learning and information retrieval. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/khalaf-2020-lebanonuprising-thorough-study-lebanese-tweets/