---

🎓 GenAI-EdSent: Sentiment-Driven Evaluation of AI Educational Apps

Unveiling User Perceptions in the Generative AI Era Through Large-Scale Sentiment Analysis

---

📄 Associated Paper

Title: Unveiling User Perceptions in the Generative AI Era: A Sentiment-Driven Evaluation of AI Educational Apps' Role in Digital Transformation of E-Teaching

Authors: Adeleh Mazaheriyan (Islamic Azad University) & Erfan Nourbakhsh (University of Isfahan)

Abstract: This study performs a sentiment-driven evaluation of user reviews from 22 top AI ed-apps on the Google Play Store to assess efficacy, challenges, and pedagogical implications. Our pipeline leverages RoBERTa for binary sentiment classification, GPT-4o for key point extraction, and GPT-5 for synthesizing top positive/negative themes. Results reveal predominantly positive sentiments, with homework apps leading (e.g., Edu AI: 95.9% positive) while specialized LMS/language apps lag due to stability issues.

---

📊 Research at a Glance

📱 Apps Analyzed	💬 Reviews Processed	🤖 AI Models Used	📚 Categories
22	481,000+	3 (RoBERTa, GPT-4o, GPT-5)	7

1️⃣ App Selection & Categorization

• Selected 22 prominent AI education apps based on ratings, downloads, and GenAI integration
• Categorized into 7 functional types with overlaps for multifunctional designs:
  • 📝 AI Quiz & Question Generators
  • 🎒 All-in-One Study Companions
  • ✏️ Homework Helpers
  • 🔢 Math-Focused Solvers & Specialized Tools
  • 📄 Document/Content Tools
  • 🏫 Learning Management Systems (LMS)
  • 🌐 Language Learning Apps

2️⃣ Data Collection

• Web scraping from Google Play Store using Python
• Collected app metadata + verbatim user reviews (up to November 2025)
• Dataset: Tens of thousands to millions of reviews per app

📊 Dataset

The full dataset containing 481,000+ app reviews from 22 AI educational apps is publicly available on Hugging Face:

🔗 GenAI-EdSent Dataset on Hugging Face

The dataset includes:

• App Information: Metadata for all 22 applications (ratings, descriptions, install counts)
• User Reviews: Complete review texts with scores, timestamps, and sentiment labels
• Sentiment Analysis Results: Binary classification outputs from RoBERTa
• Extracted Insights: Key positive/negative themes per application

📚 Citation

If you use this work in your research, please cite:

@article{mazaherian2025unveiling,
  title={Unveiling User Perceptions in the Generative AI Era: A Sentiment-Driven Evaluation of AI Educational Apps' Role in Digital Transformation of e-Teaching},
  author={Mazaherian, Adeleh and Nourbakhsh, Erfan},
  journal={arXiv preprint arXiv:2512.11934},
  year={2025}
}

📄 License

This dataset is licensed under the CC-BY-4.0 License. The code is licensed under the MIT License.

---

Made with ❤️ for the future of education