Elamur Research⌕
☰
home / papers / 2020 / Arabic Opinion Mining Using a Hybrid Recommender System Approach

Arabic Opinion Mining Using a Hybrid Recommender System Approach

Summary

The paper addresses data sparsity in recommender systems. Many users never leave star ratings, so collaborative filtering has little to work with. The authors propose a hybrid pipeline that recovers extra rating signal from what users write. They focus on Arabic, using the Opinion Corpus for Arabic (OCA) of movie reviews.

The pipeline has two stages:

  • Sentiment stage. Reviews are cleaned of stop words, special characters, punctuation and digits. Word features are extracted and pruned by feature selection, then encoded as a TFIDF matrix. An SVM labels each review as positive (+1) or negative (−1).
  • Recommendation stage. The predicted polarity is merged with the user's star rating into a single user–movie matrix. A model-based (SVD-style) collaborative filter then predicts ratings for unseen movies.

The system reaches roughly 85% accuracy in inferring ratings from reviews. It clearly beats the baseline on precision, recall and F-measure. It also lowers MAE compared with an item-based CF baseline. The work matters because it shows free-text opinions can partly fill rating gaps in a lower-resource language.

Key points

  • Data: OCA (Rushdi-Saleh et al.), 500 Arabic movie reviews gathered from several websites, split evenly into 250 positive and 250 negative.
  • Sentiment classifier: SVM over TFIDF features after feature selection.
    • Precision: 93.6% vs 46.2% for the baseline.
    • Recall: 76.6% vs 48.0%.
    • F-measure: 84.3% vs 47.1%.
  • Fusion: review polarity (±1) is combined with the star rating via the paper's Formula 3. The resulting matrix feeds a model-based collaborative filter.
  • Rating prediction (MAE vs item-based CF baseline):
    • At R=0.5: 0.523 vs 0.579.
    • At R=0.9: about 0.42 (the table and text differ slightly) vs 0.450.
    • The model wins at every train ratio from 0.1 to 0.9.
  • Stated future work: larger datasets with more users and items, and more tuning of feature selection.

Interesting ideas

  • I1. When the explicit numeric signal is sparse, add a second evidence channel derived from text and merge both into one per-user, per-item value. — For Mova (world): Entity states could keep stated values and text-derived evaluations as separate, traceable evidence sources. A combined judgment could then be computed while still explaining which source contributed what.
  • I2. Report results over a sweep of train/test ratios (R from 0.1 to 0.9) rather than a single split, which shows how the method behaves as data becomes scarce. — For Mova (learn): Parser and reader evaluations could be plotted against the amount of silver training data. This would reveal whether a gain survives in low-data regimes or only appears with plenty of data.
  • I3. The paper's example shows that the same attribute can flip polarity by domain: a quiet washing machine is good, but a quiet TV is bad. Polarity therefore depends on what the entity is for. — For Mova (frames): Evaluative frames could look up the entity's function in the ontology before assigning polarity to an attribute such as loudness. The explanation would then show why "low sound" counted as praise or criticism.
  • I4. The related work stresses keeping a neutral class, because many sentences carry no opinion at all. — For Mova (frames): The speech-act and evaluation layer should be allowed to output "no stance" explicitly rather than forcing a polarity. This reduces false sentiment attached to plain statements.

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

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.1142/s2717554520500101 · arXiv 2009.07397

@misc{harrag2020arabic,
  title = {{Arabic Opinion Mining Using a Hybrid Recommender System Approach}},
  author = {Harrag, Fouzi and Al-Salman, Abdulmalik Salman and Alquahtani, Alaa},
  year = {2020},
  doi = {10.1142/s2717554520500101},
  eprint = {2009.07397},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.1142/s2717554520500101}
}

References

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

  1. Bing Liu. 2012. Sentiment analysis and opinion mining. Synthesis Lectures on Human Language Technologies, 5(1):1–167. ↗
  2. Bing Liu. 2015. Sentiment analysis: Mining opinions, sentiments, and emotions. Cambridge University Press. K. Schouten and F. Frasincar, "Survey on Aspect-Level Sentiment Analysis", IEEE Transactions on Knowledge and Data Engineering, vol. 28, no. 3, pp. 813-830, 2016. W. Che, Y. Zhao, H. Guo, Z. Su and T. Liu, "Sentence Compression for Aspect-Based Sentiment Analysis", IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 23, no. 12, pp. 2111-2124, 2015. K. Sarawgi and V. Pathak, "Opinion Mining: Aspect Level Sentiment Analysis using SentiWordNet and Amazon Web Services", International Journal of Computer Applications, vol. 158, no. 6, pp. 31-36, 2017. S. Alhojely, "Different Applications and Techniques for Sentiment Analysis", International Journal of Computer Applications, vol. 154, no. 5, pp. 24-28, 2016. N. S. Altwairesh, “Sentiment Analysis of Twitter: A Study on the Saudi Community,” Ph.D dissertation, KSU Univ., Riyadh, 2016. M. Ayyoub, S. Essa and I. Alsmadi, "Lexicon-based sentiment analysis of Arabic tweets", International Journal of Social Network Mining, vol. 2, no. 2, p. 101, 2015 H. AL-Rubaiee, R. Qiu and D. Li, "The Importance of Neutral Class in Sentiment Analysis of Arabic Tweets", International Journal of Computer Science and Information Technology, vol. 8, no. 2, pp. 17-31, 2016. J. Ben Salamah and A. Elkhlifi, "Microblogging Opinion Mining Approach for Kuwaiti Dialect", in International conference on Computing Technology and Information Management, 2014. M. Al-Kabi, I. Alsmadi, A. Gigieh, H. Wahsheh and M. Haidar, "Opinion Mining and Analysis for Arabic Language", International Journal of Advanced Computer Science and Applications, vol. 5, no. 5, 2014. C. Aggarwal, Recommender systems, 1st ed. 2016. G. Kembellec, G. Chartron and I. Saleh, Recommender Systems, 1st ed. find
  3. John Wiley & Sons, 2015. P. Tugaonkar and V. Chitre, "Survey on recent methodologies used for recommender system", 2016 International Conference on Electrical, Electronics, and Optimization Techniques (ICEEOT), 2016. F. Isinkaye, Y. Folajimi and B. Ojokoh, "Recommendation systems: Principles, methods and evaluation", Egyptian Informatics Journal, vol. 16, no. 3, pp. 261-273, 2015. Z. Yang, B. Wu, K. Zheng, X. Wang and L. Lei, "A Survey of Collaborative Filtering-Based Recommender Systems for Mobile Internet Applications", IEEE Access, vol. 4, pp. 3273-3287, 2016. find
  4. Arabic Opinion Mining Using a Hybrid Recommender Systems Approach 17. E. Ashley-Dejo, S. Ngwira and T. Zuva, "A survey of Context-aware Recommender System and services", 2015 International Conference on Computing, Communication and Security (ICCCS), 2015. 18. P. Yadav, "Recommender System: Revolution in E-Commerce", International Journal Of Engineering And Computer Science, 2016. 19. H. Ambulkar and A. Pathan, "Recommender System Challenges and Methodologies in Social Network: Survey", International Journal of Science and Research (IJSR), vol. 4, no. 11, pp. 286-289, 2015. 20. K. Verbert, N. Manouselis, X. Ochoa, M. Wolpers, H. Drachsler, I. Bosnic and E. Duval, "Context-Aware Recommender Systems for Learning: A Survey and Future Challenges", IEEE Transactions on Learning Technologies, vol. 5, no. 4, pp. 318-335, 2012. 21. M. Lerato, O. Esan, A. Ebunoluwa, S. Ngwira and T. Zuva, "A survey of recommender system feedback techniques, comparison and evaluation metrics", 2015 International Conference on Computing, Communication and Security (ICCCS), 2015. 22. E. Alshari, A. Azman, N. Mustapha, S. Doraisamy and M. Alksher, "Prediction of rating from comments based on information retrieval and sentiment analysis",2016 Third International Conference on Information Retrieval and Knowledge Management (CAMP), 2016. 23. N. Omar and A. Eesee, "A Hybrid Method for Arabic Educational Sentiment Analysis", Journal of Applied Sciences, vol. 16, no. 5, pp. 216-222, 2016. 24. H. Mobarz, M. Rashown and I. Farag, "Using Automated Lexical Resources in Arabic Sentence Subjectivity", International Journal of Artificial Intelligence & Applications, vol. 5, no. 6, pp. 01-14, 2014. 25. G. Badaro, R. Baly, H. Hajj, N. Habash and W. El-Hajj, "A Large Scale Arabic Sentiment Lexicon for Arabic Opinion Mining", in Proceedings of the EMNLP Workshop on Arabic Natural Language Processing (ANLP), 2014, pp.165-173. 26. L. Almuqren and A. Cristea, "Framework for Sentiment Analysis of Arabic Text", Proceedings of the 27th ACM Conference on Hypertext and Social Media - HT '16, 2016. 27. A. Shoukry and A. Rafea, "Sentence-level Arabic sentiment analysis",2012 International Conference on Collaboration Technologies and Systems (CTS), 2012. 28. R. Duwairi, N. Ahmed and S. Al-Rifai, "Detecting sentiment embedded in Arabic social media – A lexicon-based approach", Journal of Intelligent & Fuzzy Systems, vol. 29, no. 1, pp. 107-117, 2015. 29. M. Abdul-Mageed, M. Diab and S. Kübler, "SAMAR: Subjectivity and sentiment analysis for Arabic social media”, Computer Speech & Language, vol. 28, no. 1, pp. 20-37, 2014. 30. Younis, M.G.E. “Sentiment Analysis and Text Mining for Social Media Microblogs using Open Source Tools: An Empirical Study”, International Journal of Computer Applications,vol. 112, no. 5, 2015 31. M. Rushdi-Saleh, M. Martín-Valdivia, L. Ureña-López and J. Perea-Ortega, "OCA: Opinion corpus for Arabic", Journal of the American Society for Information Science and Technology, vol. 62, no. 10, 2011. 32. M. Al-Smadi, O. Qawasmeh, B. Talafha and M. Quwaider, "Human Annotated Arabic Dataset of Book Reviews for Aspect Based Sentiment Analysis", 2015 3rd International Conference on Future Internet of Things and Cloud, 2015. 33. M. Aly and A. Atiya, "LABR: A Large Scale Arabic Book Reviews Dataset", ACL, 2013. 34. R. Eskander and O. Rambow, "SLSA: A Sentiment Lexicon for Standard Arabic", in Empirical Methods in Natural Language Processing, Lisbon, Portugal, 2015. 35. S. Alhumoud, T. Albuhairi and W. Alohaideb, "Hybrid sentiment analyser for Arabic tweets using R", in Knowledge Engineering and Knowledge Management (IC3K), Lisbon, Portugal, 2015. find
  5. Arabic Opinion Mining Using a Hybrid Recommender Systems Approach 36. M. Khorsheed and A. Al-Thubaity, "Comparative evaluation of text classification techniques using a large diverse Arabic dataset" in Language Resources and Evaluation, 47(2), pp.513-538, 2013. 37. X. Lei, X. Qian and G. Zhao, "Rating Prediction Based on Social Sentiment From Textual Reviews", IEEE Transactions on Multimedia, vol. 18, no. 9, pp. 1910-1921, 2016. 38. M. Vozalis and K. Margaritis, "Using SVD and demographic data for the enhancement of generalized Collaborative Filtering", Information Sciences, vol. 177, pp. 3017–3037, 2007. 39. S. Zhang, W. Wang, J. Ford, F. Makedon, and J. Pearlman, "Using singular value decomposition approximation for collaborative filtering", in International Conference on E-Commerce Technology (CEC ’05), Washington, DC, USA, pp. 257–264, 2005. 40. A. Paterek, "Improving regularized singular value decomposition for collaborative filtering", in 13th ACM International Conference on Knowledge Discovery and Data Mining, San Jose, CA, USA, pp. 39–42, 2007. 41. B. Sarwar, G. Karypis, J. Konstan, and J. Reidl, “Item-Based Collaborative Filtering Recommendation Algorithms,” Proc. 10th Int’l Conf. World Wide Web (WWW ’01), pp. 285-295, 2001. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/harrag-2020-arabic-opinion-mining-using-hybrid/