Automatic Generation of Meteorological Briefing by Event Knowledge Guided Summarization Model

Kaize Shi,Hao Lu,Yifan Zhu,Zhendong Niu

DOI: https://doi.org/10.1016/j.knosys.2019.105379

IF: 8.139

2019-01-01

Knowledge-Based Systems

Abstract:In recent years, frequent meteorological disasters have brought great suffering to people. The meteorological briefing is an effective way to realize the real-time perception of extreme meteorological events, which is of great significance for decision-makers to formulate plans and provide timely assistance. Traditional meteorological briefings primarily rely on physical sensors for data collection and are organized manually. However, such an approach has the disadvantages of rigid content, high cost, and poor real-time performance. As an emerging lightweight social sensor, social networks can respond to real world events in a timely and comprehensive manner, which also makes up for the shortcomings of the traditional methods. In this paper, we present an event knowledge guided summarization (EKGS) model to automatically summarize weibo posts in the meteorological domain. Our model consists of two modules: a summary generation module and an event knowledge guidance module. The event knowledge guidance module is used to guide and constrain the content generated by the summary generation module, so that it can generate the content with core knowledge of specific events, which are 14 types of extreme meteorological events defined by the China Meteorological Administration (CMA). Compared to other baseline models, our EKGS model achieves the best test results on all metrics. In addition, we construct an automatic meteorological briefing generation framework based on the EKGS model, which has been applied as an online service to the meteorological briefing overview module of the CMA Public Meteorological Service Center.