From Machine Learning to Heterogeneous Models: A Comprehensive Study on Fine-Grained Emotion Detection in Arabic Text
Received: 22 January 2026 | Revised: 24 February 2026 | Accepted: 4 March 2026 | Online: 4 April 2026
Corresponding author: Esra'a Alshdaifat
Abstract
Recognizing emotions from Arabic text has received increasing attention in the last few years, due to its significant role in several real-world applications. This study begins by employing several classic Machine Learning (ML), Deep Learning (DL), and pre-trained Large Language Models (LLMs) to construct emotion detection models for Arabic text. Then, a composite model is proposed that integrates the most-effective model from each considered category. More precisely, two main strategies are examined to combine the most-effective models: (i) a parallel strategy and (ii) a sequential strategy. Adopting the parallel strategy, the base models are independent and combined utilizing majority voting or average probability techniques, whereas in the sequential strategy the models are dependent and different weights are assigned to the base models. According to the extensive experiments conducted, the best F1-scores obtained from ML, DL, and pre-trained LLMs were 0.668, 0.742, and 0.794, respectively. On the other hand, the sequential composite model produced an F1-score of 0.802, outperforming the parallel models, which reached an F1-score of 0.790. The findings highlight the power of sequential hybrid models and confirm that integrating different architectural models in a way that one model passes information to the succeeding model, focusing on learning hard instances, outperforms all considered models.
Keywords:
emotion detection, Natural Language Processing (NLP), heterogeneous models, sequential hybrid modelsDownloads
References
S. Mohammad, F. Bravo-Marquez, M. Salameh, and S. Kiritchenko, "SemEval-2018 Task 1: Affect in Tweets," in Proceedings of the 12th International Workshop on Semantic Evaluation, New Orleans, LA, USA, 2018, pp. 1–17. DOI: https://doi.org/10.18653/v1/S18-1001
A. Al-Khatib and S. R. El-Beltagy, "Emotional Tone Detection in Arabic Tweets," in Computational Linguistics and Intelligent Text Processing: 18th International Conference, CICLing 2017, Budapest, Hungary, 2017, pp. 105–114. DOI: https://doi.org/10.1007/978-3-319-77116-8_8
T. Al-Moslmi, M. Albared, A. Al-Shabi, N. Omar, and S. Abdullah, "Arabic senti-lexicon: Constructing publicly available language resources for Arabic sentiment analysis," Journal of Information Science, vol. 44, no. 3, pp. 345–362, June 2018. DOI: https://doi.org/10.1177/0165551516683908
A. Assiri, A. Emam, and H. Al-Dossari, "Towards enhancement of a lexicon-based approach for Saudi dialect sentiment analysis," Journal of Information Science, vol. 44, no. 2, pp. 184–202, Apr. 2018. DOI: https://doi.org/10.1177/0165551516688143
J. Serrano-Guerrero, B. Alshouha, F. P. Romero, and J. A. Olivas, "Affective Knowledge-enhanced Emotion Detection in Arabic Language: A Comparative Study," JUCS - Journal of Universal Computer Science, vol. 28, no. 7, pp. 733–757, July 2022. DOI: https://doi.org/10.3897/jucs.72590
G. Badaro et al., "EMA at SemEval-2018 Task 1: Emotion Mining for Arabic," in Proceedings of the 12th International Workshop on Semantic Evaluation, New Orleans, LA, USA, 2018, pp. 236–244. DOI: https://doi.org/10.18653/v1/S18-1036
A. B. Soliman, K. Eissa, and S. R. El-Beltagy, "AraVec: A set of Arabic Word Embedding Models for use in Arabic NLP," Procedia Computer Science, vol. 117, pp. 256–265, Jan. 2017. DOI: https://doi.org/10.1016/j.procs.2017.10.117
P. Bojanowski, E. Grave, A. Joulin, and T. Mikolov, "Enriching Word Vectors with Subword Information," Transactions of the Association for Computational Linguistics, vol. 5, pp. 135–146, June 2017. DOI: https://doi.org/10.1162/tacl_a_00051
H. Mulki, C. Bechikh Ali, H. Haddad, and I. Babaoğlu, "Tw-StAR at SemEval-2018 Task 1: Preprocessing Impact on Multi-label Emotion Classification," in Proceedings of the 12th International Workshop on Semantic Evaluation, New Orleans, LA, USA, 2018, pp. 167–171. DOI: https://doi.org/10.18653/v1/S18-1024
F. A. Q. Aljradi, M. Albared, A. S. Ghareb, and A. Thawaba, "An Ensemble Framework for Imbalanced Arabic Text-Based Emotion Analysis," Journal of Science and Technology, vol. 30, no. 1, pp. 60–71, 2025. DOI: https://doi.org/10.20428/jst.v30i1.2566
M. Baali and N. Ghneim, "Emotion analysis of Arabic tweets using deep learning approach," Journal of Big Data, vol. 6, no. 1, Oct. 2019, Art. no. 89. DOI: https://doi.org/10.1186/s40537-019-0252-x
N. Alswaidan and M. E. B. Menai, "Hybrid Feature Model for Emotion Recognition in Arabic Text," IEEE Access, vol. 8, pp. 37843–37854, 2020. DOI: https://doi.org/10.1109/ACCESS.2020.2975906
A. J. Almahdawi and W. J. Teahan, "A New Arabic Dataset for Emotion Recognition," in Proceedings of the 2019 Computing Conference, Volume 2, London, UK, 2019, pp. 200–216. DOI: https://doi.org/10.1007/978-3-030-22868-2_16
E. A. H. Khalil, E. M. F. E. Houby, and H. K. Mohamed, "Deep learning for emotion analysis in Arabic tweets," Journal of Big Data, vol. 8, no. 1, Oct. 2021, Art. no. 136. DOI: https://doi.org/10.1186/s40537-021-00523-w
A. Tolep, S. Mamikov, Z. Yakhiya, and A. Abdullayeva, "Sentiment Analysis of Social Network Data Using Traditional and Hybrid Deep Learning Models," Engineering, Technology & Applied Science Research, vol. 15, no. 6, pp. 28481–28488, Dec. 2025. DOI: https://doi.org/10.48084/etasr.12865
M. J. Althobaiti, "Arabic Emotion Recognition in Low-Resource Settings: A Novel Diverse Model Stacking Ensemble with Self-Training," Applied Sciences, vol. 13, no. 23, Nov. 2023, Art. no. 12772. DOI: https://doi.org/10.3390/app132312772
A. El-Sayed, M. Abougabal, and S. Lazem, "Benchmarking a large Twitter dataset for Arabic emotion analysis," SN Applied Sciences, vol. 5, no. 8, July 2023, Art. no. 224. DOI: https://doi.org/10.1007/s42452-023-05437-1
M. Abdul-Mageed, A. Elmadany, and E. M. B. Nagoudi, "ARBERT & MARBERT: Deep Bidirectional Transformers for Arabic," in Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), Online, 2021, pp. 7088–7105. DOI: https://doi.org/10.18653/v1/2021.acl-long.551
A. Abdelali, S. Hassan, H. Mubarak, K. Darwish, and Y. Samih, "Pre-Training BERT on Arabic Tweets: Practical Considerations." arXiv, Feb. 21, 2021.
N. Al-Twairesh, "The Evolution of Language Models Applied to Emotion Analysis of Arabic Tweets," Information, vol. 12, no. 2, Feb. 2021, Art. no. 84. DOI: https://doi.org/10.3390/info12020084
A. Safaya, M. Abdullatif, and D. Yuret, "KUISAIL at SemEval-2020 Task 12: BERT-CNN for Offensive Speech Identification in Social Media," in Proceedings of the Fourteenth Workshop on Semantic Evaluation, Barcelona, Spain (Online), 2020, pp. 2054–2059. DOI: https://doi.org/10.18653/v1/2020.semeval-1.271
F. Qarah, "EgyBERT: A Large Language Model Pretrained on Egyptian Dialect Corpora." arXiv, Aug. 07, 2024.
A. Conneau et al., "Unsupervised Cross-lingual Representation Learning at Scale," in Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, Online, 2020, pp. 8440–8451. DOI: https://doi.org/10.18653/v1/2020.acl-main.747
Downloads
How to Cite
License
Copyright (c) 2026 Esra'a Alshdaifat, Enshirah Altarawneh, Mogeeb Alrahman Aloshaibat, Anas Hussein, Mohammed Dawood, Tawfiq Tahaineh
This work is licensed under a Creative Commons Attribution 4.0 International License.
Authors who publish with this journal agree to the following terms:
- Authors retain the copyright and grant the journal the right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) after its publication in ETASR with an acknowledgement of its initial publication in this journal.
