Learning Event Patterns from Text
Mitja Trampus and Dunja Mladenić
We propose a pipeline for learning event templates from a large corpus of textual news articles. An event template is a machine-usable semantic data structure, in our case a graph, describing a certain event type. Such a template encodes the most characteristic information for a certain type of event; for instance, an earthquake template would encode "x people dead” and/or “town y shook at time z". Event templates have potential to be used as an input for information extraction tasks or automated ontology extension. We present preliminary results of applying the proposed pipeline on a subset of news articles.