Generic Event Extraction Using Markov Logic Networks

Abstract

In this thesis, we will fi rst explore extensively the state-of-the-art research of event extraction. Then we will present our framework in MLNs to solve the event extraction task as defi ned in the Automatic Content Extraction (ACE) Program. Finally, we will demonstrate how to extend our framework from sentence level to document level and how to incorporate document-level features, like event correlation information, into our framework. We conducted extensive experiments on the ACE 2005 English corpus, to evaluate the generic event extraction scenario. Experimental results show that our system is both e ffcient and e ffective in extracting events from text documents. Our framework could make use of the joint learning function provided by MLNs, thus the error propagation problem which is severe and occurs frequently in pipeline systems can be easily avoided. Finally, we have achieved statistically signi ficant improvement after incorporating event correlation information into our framework.