Typhoon disaster knowledge service driven by large language models: Key technologies and applications

This study is led by researcher Yi Huang and Yongqi Xia (School of Internet of Things, Nanjing University of Posts and Telecommunications), researcher Xueying Zhang and Yehua Sheng (Key Laboratory of Virtual Geographic Environment, Nanjing Normal University, Ministry of Education), and researcher Peng Ye (Urban Planning and Development Institute, Yangzhou University). The researchers first introduce a comprehensive framework for typhoon disaster knowledge services, comprising three interconnected layers as data, knowledge, and service.

 

At the data-to-knowledge layer, this study proposes an LLM-driven automated method for constructing typhoon disaster knowledge graphs (KGs). This method employs a "pre-training + fine-tuning" paradigm to transform unstructured textual data into a structured knowledge representation. A specialized training dataset was then curated, incorporating typhoon-related texts with explicit temporal and spatial attributes. The resulting KG integrates 7,221 nodes, 52 edges, and 5,776 triplets, capturing spatiotemporal dynamics and disaster impact mechanisms.

 

The aforementioned structured knowledge forms a foundation for the subsequent service layer, where an LLM-based intelligent question-answering system is developed. Utilizing Graph Retrieval-Augmented Generation (GraphRAG), the system retrieves contextually relevant knowledge from the KG and generates user-specific guidance, such as evacuation plans and resource allocation strategies. The authors demonstrate the effectiveness of this system by substantial experiments, which achieves high accuracy in concept, entity, attribute, and relation extraction. Such knowledge-to-service layer ensures seamless conversion of structured knowledge into practical services, such as personalized evacuation plans or resource allocation strategies.

 

To sum up, this study highlights the transformative potential of LLMs in typhoon disaster management and lays a foundation for integrating LLMs with geospatial technologies. For more details, please refer to the original article:

Typhoon Disaster Knowledge Service Driven by Large Language Models: Key Technologies and Applications

https://www.sciengine.com/JGIS/doi/10.12082/dqxxkx.2025.250175 （If you want to see the English version of the full text, please click on the 科大讯飞翻译(iFLYTEK Translation)   in the article page.）

 

---

---