Event Nugget Detection Using Thresholding and Classification Techniques.

This paper describes our Event Nugget Detection system that we submitted to the TAC KBP 2015. We considered the problem as document classification problem. We used confidence scores from classifier to detect the event by thresholding. Our feature vectors consist of event nugget context, part of speech tags of event nugget context and semantic similarity score between event nugget and event subtypes. Our performance was low; we got F1 measure of 0.33 for both event nugget detection task and coreference task.