EXTRACTION AND APPLICATION OF GEOLOCALIZED DICTIONARIES

References: 14th International Multidisciplinary Scientific GeoConference SGEM 2014, www.sgem.org, SGEM2014 Conference Proceedings, ISBN 978-619-7105-10-0 / ISSN 1314-2704, June 19-25, 2014, Book 2, Vol. 1, 593-600 pp

ABSTRACT
The field of natural language processing gains more and more in popularity, especially in the context of the growing demand for discovery of valuable content in continuously growing human generated data. This paper presents a method of geolocalized information retrieval from text with special concern placed on automatic recognition of semantic geolocalized relations, their classification and interpretation resulting in assigning them specific geographical coordinates. The semantic relations are matched against a geolocalized dictionary based on the OpenStreetMap project and stored in a spatial database created by extending PostgreSQL object-relational database with the PostGIS module. A proof of concept implementation is discussed, using the data stored in the aforementioned spatial database. Its performance according to diverse metrics is presented. This paper focuses on determining the coordinates of locations in Cracow (Poland) and a spatial reference system (SRS) appropriate to this area was used. However, the presented methods and implemented system may be applied in any other place in the World and use any spatial reference system. The natural language texts used in the described experiment were in Polish and coming from a real life application being a part of the IBM Smart Cities initiative. The lexical challenges added an additional layer of complexity due to the complex inflectional system of the source language. But it has to be emphasized again that the discussed theory and method of assigning geographical coordinates to a textually described locations, alongside with its implementation, is applicable to any other language.

Keywords: geolocalization, natural language processing, geolocalized relations, geolocalized dictionaries