一种基于小语言模型的城市内涝数据产品生成方法及装置
The method for generating urban flooding data products based on a small language model solves the problems of low data quality and high cost in existing technologies for emergency response to urban rainstorms and floods. It achieves high-accuracy and low-cost structured data generation, and the generated GIS data products have high practical value.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- UNIV OF SCI & TECH BEIJING
- Filing Date
- 2026-01-13
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies struggle to quickly and accurately generate high-quality structured urban flooding data products in emergency responses to urban rainstorms and floods. They suffer from problems such as sparse monitoring points, delayed information timeliness, high data noise, difficulty in identifying semantic ambiguities, high costs, and data security risks.
A method for generating urban flooding data products based on a small language model is adopted. An urban flooding data product generation device is constructed through data acquisition, fine-tuning, filtering and spatial deduplication modules. The device includes a data acquisition module, a dataset construction module, a fine-tuning module, a filtering module and a data structuring module. LoRA and 4-bit NF4 quantization techniques are used for model fine-tuning. Combined with the Haversine distance formula and semantic priority strategy, high-purity structured data is generated.
It achieves efficient conversion from unstructured crowdsourced text to structured data, significantly improves the accuracy of water accumulation point identification, reduces costs, ensures data security, and generates GIS data products with high practical value.
Smart Images
Figure CN121960487B_ABST