Mining Wikipedia Article Clusters for Geospatial Entities and Relationships

No. 8: Social Semantic Web: Where Web 2.0 Meets Web 3.0

Volume

Papers from the 2009 AAAI Spring Symposium

Contents

We present in this paper a method to extract geospatial entities and relationships from the unstructured text of the English language Wikipedia. Using a novel approach that applies SVMs trained from purely structural features of text strings, we extract candidate geospatial entities and relationships. Using a combination of further techniques, along with an external gazetteer, the candidate entities and relationships are disambiguated and the Wikipedia article pages are modified to include the semantic information provided by the extraction process. We successfully extracted location entities with an F-measure of 81%, and location relations with an Fmeasure of 54%