Geographical Named Entity Recognition from Travel Articles

Abstract

Geographical Named Entity Recognition, also known as geoparsing, is the task of obtaining geographical coordinates from free-format text, arises in many real-world applications such as understanding location instructions in auto-response systems, determining a document’s geographic scope, real-time social media geographical event analysis, and more. Geoparsing consists of two parts: toponym extraction from the text; toponym resolution, disambiguating and connecting toponyms to fully specified real-world locations, geographical coordinates. In this work, I tackle the problem of geoparsing travel guide articles. For this reason, I developed the geoparsing system. As input data, National Geographic Travel articles describing road trips in the United States were used. The code is available on Github[23].