Improving automated crisis detection via an improved understanding of crisis language: Linguistic categories in social media crises

Abstract

AbstractBy applying the Linguistic Category Model (LCM) in crisis communication, this study explores the potential of verb tracking on social media to examine how linguistic categories can elucidate the intentional and/or unintentional communication of crisis attribution frames. Through a content analysis, linguistic categories used in both media posts reporting three clusters of crisis and public comments on Facebook were examined. Results indicated that linguistic abstraction in both media post and public comments describing the crisis varied based on crisis cluster, suggesting that the level of linguistic abstraction reflected perceived attribution of responsibility through stability, locus and controllability. Language used to describe preventable crisis tend to be more abstract than those used to describe accidental and victim crisis. Findings of this study empirically tested the integration of LCM in crisis communication and implied potential application of LCM in building automated environmental scanning and crisis prediction systems.