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