Supervised Event Coding From Text Written in Spanish

Abstract

Recent innovations in conflict and computer research favor generating massive event data using automated coding protocols. Unfortunately, these approaches almost exclusively rely on English-language sources, thus causing problems of coverage bias and misleading inferences. In an effort to attenuate Anglocentrism in event data, we introduce Eventus ID, new software for supervised event coding from text written in Spanish. Drawing on real news reports, the application generates daily georeferenced data on how the military fights organized criminals in Mexico. Performance metrics show that Eventus ID is almost as accurate as humans for coding event data.