A Multilingual Hybrid Deep Learning Framework for Sentiment-Driven Customer Satisfaction Prediction using CNN-BGRU-LSTM and Explainable AI
Description
RateFlow is an intelligent multilingual sentiment analysis platform developed to transform customer reviews and online feedback into valuable business insights using artificial intelligence techniques. The system is designed to help organizations automatically understand customer opinions, detect satisfaction trends, and improve decision-making processes. It uses a hybrid deep learning architecture that combines CNN and LSTM/BGRU models to achieve high sentiment classification accuracy across different languages, dialects, and writing styles. In addition, the platform integrates explainable AI through SHAP, allowing users to understand how specific words or phrases influenced each prediction. Experimental evaluation demonstrated strong performance, confirming the effectiveness of hybrid models for accurate and real-time sentiment analysis in practical environments.
This work was conducted at Arab International University (AIU), Syria. The official website of the university is: https://www.aiu.edu.sy
Files
LaithAlhalabi_2026_RateFlow.pdf
Files
(1.6 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:20b08c8f0ece7c5bdf8ce8e14d85d439
|
1.6 MB | Preview Download |
Additional details
References
- [1] A. G. Al-Anzi et al., "A hybrid deep learning model for consumer sentiment analysis,".[Online].Available: https://journals.plos.org/plosone/article/figure/image?size=medium&id=10.1371/journal.pone.0294701.g004. [Accessed: April 28, 2026]
- [2] "SHAP (Shapley Additive exPlanations) summary plot visualization for model interpretability,"GitHub,2023.[Online].Available: https://user-images.githubusercontent.com/68812855/229953625-5bb289b2-10e5-4f8d-bbdf-7c2448c49ea2.png. [Accessed: April 28, 2026]
- [3] M. A. Haq, "A Hybrid Deep Learning Approach for Enhanced Sentiment Classification in Customer Reviews," Mathematics, vol. 12, no. 23, p. 3856, Dec. 2024, doi: https://doi.org/10.3390/math12233856
- [4] S. S. Aljameel et al., "Attention-augmented hybrid CNN-LSTM for social media sentiment analysis," Scientific Reports, vol. 15, no. 1, Art. no. 18245, Jan. 2025 doi: https://doi.org/10.1038/s41598-025-18245-x
- [5] J. Wang, "Hybrid Sentiment Analysis Model for Movie Reviews Based on BERT and CNN," Proceedings of the 2nd International Conference on Computing and Artificial Intelligence (ICCAI 2024), ACE - Applied Computing and Engineering, vol. 35, pp. 112-118, 2025, doi: https://doi.org/10.54254/2755-2721/2025.AST27216
- [6] R. Kumar and A. Singh, "Context-based sentiment analysis with BiGRU + DistilBERT + XGBoost for COVID-19 tweets," Scientific Reports, vol. 15, no. 1, Art. no. 22929, Feb. 2025, doi: https://doi.org/10.1038/s41598-025-22929-9
- [7] J. Doe and R. Smith, "Multilingual Sentiment Analysis in E-commerce using T5-CapsNet Ensemble," Journal of Open Innovation: Technology, Market, and Complexity,vol.11,no.1,p.100042,2025,doi: https://doi.org/10.13140/RG.2.2.26048.65285
- [8] A. Al-Haj et al., "Opinion Mining with Hybrid BGRU-LSTM for Large-Scale Reviews,"arXiv,preprint,arXiv:2511.14796,2025,doi: https://doi.org/10.3390/technologies13050175
- [9] T. Barhoum, "SETF: A Structured Engineering Thesis Framework for Artificial Intelligence, Software Engineering, and Robotics", Zenodo, Apr. 2026. doi: https://doi.org/10.5281/zenodo.19686845
- [10] DevOpsSchool, "CNN architecture for NLP and text classification," May 2023. [Online]. Available: https://www.devopsschool.com/blog/wp-content/uploads/2023/05/image-227.png. [Accessed: April 28, 2026]
- [11] M. Nguyen, "Illustrated Guide to LSTM's and GRU's: A step by step explanation," Medium, Sep. 2018. [Online]. Available: http://miro.medium.com/v2/1*QiLYco0hB8EterWYTyyv4g.png. [Accessed: April 28, 2026]
- [12] J. Wu et al., "A hybrid model based on BGRU and attention mechanism," Information Sciences, Feb. 2021. [Online]. Available: https://ars.els-cdn.com/content/image/1-s2.0-S0020025520311981-gr5.jpg. [Accessed: April 28, 2026]
