Master of Disaster: A Disaster-Related Event Monitoring System From News Streams

TL;DR

The paper addresses the challenge of discriminating real-world disaster events across noisy news streams. It introduces Master of Disaster (MoD), a semi-automatic pipeline that ingests GDELT news, extracts structured event data with a transformer-based detector and KG grounding to Wikidata via the CoyPu ontology, and visualizes embeddings with PCA-DBSCAN to aid event-instance discrimination in a human-in-the-loop setting. Key contributions include an end-to-end, open-source architecture with a data preprocessor, transformer-based event extractor, KG-grounded RDF output, and a visualization module that supports rapid human screening. The approach enables interpretable disaster-event monitoring with grounding to a knowledge graph, facilitating downstream validation and reasoning, and lays groundwork for temporal analysis and extractor performance improvements.

Abstract

The need for a disaster-related event monitoring system has arisen due to the societal and economic impact caused by the increasing number of severe disaster events. An event monitoring system should be able to extract event-related information from texts, and discriminates event instances. We demonstrate our open-source event monitoring system, namely, Master of Disaster (MoD), which receives news streams, extracts event information, links extracted information to a knowledge graph (KG), in this case Wikidata, and discriminates event instances visually. The goal of event visualization is to group event mentions referring to the same real-world event instance so that event instance discrimination can be achieved by visual screening.

Figures (2)

• Figure 1: Overview over MoD's Architecture.
• Figure 2: Event Visualization with respect to the classification (left) and clustering (right) Result. The search keyword is "Hamburg", and the date is "11. March 2023"