Using Semi-Supervised Learning and Wikipedia to Train an Event Argument Extraction System

The paper presents a methodology for training an event argument extraction system in semi-supervised setting. We use Wikipedia and Wikidata to automatically obtain small noisily labeled dataset large unlabeled dataset. consists of clusters containing pages multiple languages. data is iteratively using learning combined with probabilistic soft logic infer the pseudo-label each example from predictions base learners. proposed applied about earthquakes terrorist attacks cross-lingual Our experiments show improvement results when methodology. achieves F1-score 0.79 only used, 0.84 trained according logic.