Generating textual emergency plans for unconventional emergencies — A natural language processing approach. (April 2023)

• Record Type: Journal Article
• Authors: Ni, Weijian; Shen, Quanle; Liu, Tong; Zeng, Qingtian; Xu, Lingzhe
• Is Part Of: Safety science. Volume 160(2023)
• Publication Date: 2023-04
• Subjects: Emergency plan -- Unconventional emergency -- Natural language processing -- Deep learning -- Natural language generation; Industrial accidents -- Periodicals; Accident Prevention -- Periodicals; Safety -- Periodicals; Travail -- Accidents -- Périodiques; 363.11
• DOI: 10.1016/j.ssci.2022.106047
• Languages: English
• ISSNs: 0925-7535

Abstract

Abstract: An emergency plan is an emergency administrative document that specifies the course of actions taken to minimize the effects of a crisis or incident. Establishing high-quality emergency plans has been a fundamental task for various emergency administrative agencies. Traditionally, emergency plans are developed based on the experiences of handling past emergencies, thus may not be well applied to unconventional emergencies that arise in an unrepeatable and unpredictable manner. This work proposes a novel emergency plan generation approach to assist decision-making under unconventional emergent situations. This goal is achieved by leveraging deep-learning-based natural language techniques to explore the interrelationship between existing emergency plans developed for common emergencies and the target unconventional emergency. In particular, an emergency response knowledge base is constructed based on a large number of existing emergency plans, and the relevant part with respect to the target unconventional emergency is retrieved. Then the new emergency plan is formed by organizing the relevant knowledge guided by a pre-defined emergency plan template. Furthermore, a novel emergency plan evaluation approach is proposed to perform a comprehensive evaluation of the quality of generated emergency plans. Empirical results on a real-world unconventional emergency case verify the feasibility of our emergency plan generation approach. Highlights: Textual emergency plans are more practically usable by emergency administrative. An automatic approach to generating textual emergency plans is proposed. An emergency response knowledge base is constructed based on existing emergency plans. A new emergency plan is formed by structurally organizing the relevant knowledge. Generated emergency plan is evaluated by fuzzy linguistic group decision-making.