Multi-KGS: Generating Social Network-Based Meteorological Decision Reports Fusing With Multiple Knowledge

The increasing prevalence of meteorological disasters necessitates advanced spatiotemporal data analytics to enhance emergency response in smart cities. Social networks, as real-time crowdsourcing sensors, provide critical data streams that Generative AI (GenAI) can fuse and summarize to generate comprehensive meteorological decision reports for enhanced emergency response during abrupt weather crises. This paper introduces a Multiple Knowledge Guided Summarization (Multi-KGS) model designed to generate meteorological decision reports by fusing posts from Sina Weibo. Specifically, the Multi-KGS model comprises a summary generation module and a multiple knowledge guidance module. The summary generation module synthesizes the content of the decision report, while the multiple knowledge guidance module steers and constrains the summarization process using knowledge of meteorological events and geographical locations, ensuring that the generated report highlights the core knowledge expressed in the source posts. Compared to baseline models, Multi-KGS achieves superior performance in content evaluation, as measured by ROUGE-1, ROUGE-2, and ROUGE-L, as well as in sentiment evaluation, with the best F1 score. This study provide a generative decision support paradigm for servicing urban computing.