Linking Events and Locations in Political Text

Abstract

This work proposes the first general technique for automatically geolocating political events in text as one step in a broader project of recognizing political relationship and extracting political information from text. Using techniques borrowed from computational linguistics and a novel set of labeled sentences, I create an method to link events and locations in text that triples the accuracy of a reasonable baseline model. I describe the potential uses of such a system in political science, describe the neural-net based model, and demonstrate its ability to answer an open question on the role of conventional military offensives in causing civilian casualties in the Syrian civil war.