TourismMinds: A Geo-augmented LLM Framework for Semantic-aware Trajectory Analytics and Generation

A groundbreaking LLM-driven framework, TourismMinds, achieves state-of-the-art performance in semantic-aware trajectory analytics and generation for tourism. By integrating GPS, social media, POI, and weather data through a five-stage spatiotemporal-semantic alignment pipeline, the model reduces hallucinations by 15% and boosts geographic reasoning accuracy. It introduces geo-spatially aware LoRA fine-tuning and hierarchical Retrieval-Augmented Generation (RAG), enhancing intent recognition by 15% and landmark identification to 72% accuracy. The framework also demonstrates 89% similarity in generating spatially plausible, personalized trajectories compared to real-world data, even in data-scarce settings. Validated using Kulangsu's smart tourism system, this approach offers a scalable solution for personalized services and decision-making in digital tourism.

Ye, Z., Su, X., He, D., Zhang, Z., Sun, Y., Ji, J., Xie, X., Wang, J., Wei, R., Zhang, Y., Li, Y., & Chen, L. (2025). TourismMinds: A Geo-augmented LLM Framework for Semantic-aware Trajectory Analytics and Generation. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 9(4), 1–32. https://doi.org/10.1145/3770697