A PDDL based Disaster Storyline Generation Approach

The rapid development of computer technology has brought challenges to the management of disaster information. When an emergency occurs, a huge amounts of relevant information will be produced. In order to form a clear event development context from the overloaded Internet data and improve the corresponding efficiency of emergency events, in this paper, we propose a method for disaster storyline construction based on intelligent planning. Specifically, we first extract the semantic information from the News text to build a Planning Domain Definition Language(PDDL) domain model, which contains complete domain knowledge. Then we use a planner to plan the event storylines that meets the presets in the domain model. To our knowledge, this is the first attempt to combine natural language processing(NLP) technology with intelligent planning technology for disaster storyline generation. Different from the previous method, our proposed model is more focused both on the event evolve information and the government emergency measures, which can provide more reference value for relevant emergency response.