Joint Extraction Model of Entities and Events

Abstract

. Joint extraction of entities and events is an important task in information extraction. In order to obtain entities and events in the text simultaneously, in this paper we firstly propose a novel tagging scheme that can transform the joint extraction task to a tagging problem. Then, based on our tagging scheme, we use different end-to-end models to extract entities and events directly and we also propose an improved objective function with different parameters to express the importance of different labels. We conduct experiments on a financial dataset and the results show that our methods are better than other existing models.