AUTHOR:

Filipe Rodrigues (fmpr [at] dei.uc.pt)

ABSTRACT:

With the current broad use of location aware devices, the activity of geo-tagging is becoming normal. The most atomic unit of this activity is the Point Of Interest (or landmark) which consists of a pair of Latitude/Longitude coordinates and a tag, normally the name of the place, a word that unambiguously identifies it and, possibly, some extra information such as the category of the POI. We present a solution to automatically extract POIs from various sources on the Web, such as Yahoo, Manta and Yellow Pages, aggregating them in a large database that can be used in navigation, characterization of a place, land use analysis and geo-reference of texts. This solution is also able to detect equivalent POIs between the multiple sources and to automatically classify them to a widespread taxonomy like the North American Industry Classification System (NAICS). We also propose a solution to automatically infer places of interest based on geo-referenced content available on the Internet like geo-tagged photos, blog posts and news feeds, thus giving a different perspective of the city and reducing the dependency on large POI directories.