计算机科学
应急响应
灾害应对
人工智能
语言模型
组分(热力学)
动作(物理)
鉴定(生物学)
自然语言
钥匙(锁)
应急程序
作者
Mingwei Tang,Yunxuan Xie,Xianggui Tan
出处
期刊:
[Institute of Electrical and Electronics Engineers]
日期:2026-05-01
卷期号:: 402-411
标识
DOI:10.1109/isssr69973.2026.00057
摘要
Emergency response is a complex, timecritical task that relies on professional expertise and practical experience. However, under time pressure, it is a challenge for humans to apply emergency knowledge flexibly and make effective decisions. Large language models (LLMs) offer a potential solution to this challenge. Nevertheless, general LLMs lack professional emergency domain training and cannot well adapt to complex and changeable emergency scenarios. To address this limitation, this study finetuned a local LLM via the Low-Rank Adaptation (LoRA) method using contingency plan data. Based on it, an intelligent emergency response approach is proposed. The method enhances the model's emergency knowledge with incident case data and applies text embedding to improve solution retrieval efficiency and accuracy. An experiment involving 6000 emergencyrelated questions was conducted to evaluate the validity of the proposed approach using the method of Fuzzy Comprehensive Evaluation (FCE). The results show that the proposed approach significantly improves the generation of emergency response solutions and can provide effective responses for specific incidents.
科研通智能强力驱动
Strongly Powered by AbleSci AI