CrisisLTLSum

CrisisLTLSum是由密歇根州立大学等机构创建的，目前最大的本地危机事件时间线数据集，包含1000个事件时间线，涵盖野火、本地火灾、交通和风暴四个领域。数据集通过半自动化流程收集，首先从Twitter流中提取集群，然后通过人工标注精炼，确保数据的质量和相关性。CrisisLTLSum主要用于支持危机管理中的实时事件跟踪和摘要生成，帮助应急响应人员快速获取和理解关键信息，有效部署资源。

CrisisLTLSum is the largest local crisis event timeline dataset to date, created by Michigan State University and other institutions. It contains 1,000 event timelines covering four domains: wildfires, local fires, traffic incidents, and storms. The dataset is collected via a semi-automated pipeline: clusters are first extracted from Twitter streams, then refined through manual annotation to ensure data quality and relevance. CrisisLTLSum is mainly designed to support real-time event tracking and summary generation in crisis management, helping emergency responders quickly obtain and understand key information and effectively deploy resources.