Corpus-Level Evaluation for Event QA: The IndiaPoliceEvents Corpus Covering the 2002 Gujarat Violence
This provides a resource for corpus-level evaluation in social science event extraction, though it is incremental as it adapts existing methods to a new domain-specific dataset.
The researchers tackled the problem of automated event extraction for social science applications by creating the IndiaPoliceEvents corpus covering the 2002 Gujarat violence, which contains 21,391 sentences from 1,257 articles with annotations for police activity events, and they evaluated off-the-shelf BERT-based models on three tasks including sentence classification and document ranking.
Automated event extraction in social science applications often requires corpus-level evaluations: for example, aggregating text predictions across metadata and unbiased estimates of recall. We combine corpus-level evaluation requirements with a real-world, social science setting and introduce the IndiaPoliceEvents corpus--all 21,391 sentences from 1,257 English-language Times of India articles about events in the state of Gujarat during March 2002. Our trained annotators read and label every document for mentions of police activity events, allowing for unbiased recall evaluations. In contrast to other datasets with structured event representations, we gather annotations by posing natural questions, and evaluate off-the-shelf models for three different tasks: sentence classification, document ranking, and temporal aggregation of target events. We present baseline results from zero-shot BERT-based models fine-tuned on natural language inference and passage retrieval tasks. Our novel corpus-level evaluations and annotation approach can guide creation of similar social-science-oriented resources in the future.