User eXperience Perception Insights Dataset (UXPID)

The UXPID dataset is a publicly available resource designed to support research in user requirements engineering, user experience (UX) analysis, and AI-driven feedback processing, particularly in scenarios where privacy or licensing may limit access to real-world user data. It enables the development and benchmarking of transformer-based models for issue detection, sentiment analysis, and requirements extraction, focusing on discussions from technical forums. UXPID fills a gap in the availability of comprehensive, annotated user feedback datasets, driving advancements in natural language processing (NLP) for industrial product support and software engineering. The dataset serves both qualitative insight extraction and large-scale machine learning applications.

Dataset Contents:

• Format: 7130 JSON files, each representing a forum discussion branch
• Comments: Over 36K synthetic user comments (mean: 5.1 comments per branch; min: 1, max: 54)
• Branches: 7130 total branches:
  • Solved branches: 2598
  • Unsolved branches: 4532
  • Question-originated branches: 7030
  • Replies: Total of 29K replies
• Annotations: Each branch and comment is annotated with information such as severity, sentiment, type, topic status, and more

This dataset is a valuable asset for researchers and practitioners looking to operationalize customer experience in technical domains, providing a solid foundation for NLP experiments and UX insight extraction where user privacy and data access are concerns.

@article{kulyabin2025user,
  title={User eXperience Perception Insights Dataset (UXPID): Synthetic User Feedback from Public Industrial Forums},
  author={Kulyabin, Mikhail and Joosten, Jan and Pacheco, Nuno Miguel Martins and Ries, Fabian and Petridis, Filippos and Bosch, Jan and Olsson, Helena Holmstr{\"o}m and others},
  journal={arXiv preprint arXiv:2509.11777},
  year={2025}
}