Cross-Context News Corpus for Protest Event-Related Knowledge Base Construction

Abstract We describe a gold standard corpus of protest events that comprise various local and international English language sources from countries. The contains document-, sentence-, token-level annotations. This facilitates creating machine learning models automatically classify news articles extract event-related information, constructing knowledge bases enable comparative social political science studies. For each source, the annotation starts with random samples continues drawn using active learning. Each batch is annotated by two scientists, adjudicated an supervisor, improved identifying errors semi-automatically. found possesses variety quality are necessary to develop benchmark text classification event extraction systems in cross-context setting, contributing generalizability robustness automated processing systems. reported results will establish common foundation collection studies, which currently lacking literature.