Chemical, biological, radiological and nuclear event detection and classification using ontology interrogation and social media data

Abstract

In an era where chemical, biological, radiological, and nuclear (CBRN) incidents present a grave threat to public safety, timely and accurate information is paramount. The complexity of the CBRN concept encompasses a range of incidents, each with unique and overlapping symptoms, related substances, and event descriptions. This study introduces an innovative approach to the development of a CBRN-specific ontology, uniting diverse data sources and domain expertise to construct a comprehensive repository of CBRN events, sub-events, their causes, symptoms, and toxic substances. Unlike prior methodologies reliant on keyword searches and predefined categories, our approach enables a holistic analysis of textual data by capturing intricate relationships between symptoms and toxic substances. We leverage this ontology in conjunction with a tailored interrogation algorithm to detect potential CBRN incidents through social media data. The algorithm was then tested on datasets of three actual CBRN incidents, one fictional incident (TV show) that simulated a nuclear incident and one non-CBRN. The interrogation algorithm was able to detect the five CBRN incidents accurately. However, the study showcased the need to extend the algorithm to distinguish between real and fictional CBRN incidents. These findings underscore the potential of this approach to deliver timely information on potential CBRN incidents. Nevertheless, the study acknowledged the inherent challenges and limitations in utilizing social media data, including the risk of misinformation, fictional events, fake news, and interference from malicious actors, all of which can affect the accuracy and reliability of the information collected.