Geonames ontology v2.0 review 3/3
Geonames document features
The classes not showed in figure 1 are drawn in figure 2. A diagram with documentation classes has been built.
The parent class is foaf Document, which means a related document that describes the feature. It has three subclasses, ontology:Map (a geonames web map), ontology:WikipediaArticle, and ontology:RDFData. So each feature can be describe by a geonames map web page, or a Wikipedia article, or RDFData respectively.
Perhaps the diferent properties defined here (ontology:locationMap, ontology:WikipediaArticle, etc.) could have been simplified. Anyway the functionality is more than satisfactory. There exist an increasing number of geonames features that are described by wikipedia articles, and an increasing number of wikipedia articles that are geocoded with geonames services.
Conclussions
The geonames ontology has nice properties for an open gazetter. It remains OWL - Lite, so it is supposed to be decidable. In the other hand, some applications could require more expresivity.
A great work has been made to define a detailed taxonomy, using SKOS philosophy. But perhaps for more complex uses, a simple taxonomy could not be enough. We could want to make special inferences about water bodies, or buildings, for example. The country class could be convenient but it is hardly elegant.
One important limitation is the geometric support, restricted to simple WGS84 points. One obvious extension should be the definition of other geometric shapes, as boxes or polygons.
The documentation part of the ontology also results confusing. Should we define a class for each web, map or entity that can document a geospatial feature?
Geonames ontology starts to deal with topological relationships. Although more topological relations could be set up, and also more complex definitions (using OWL - DL), this also could increase the ontology complexity. A compromise must be made.
More information on Geonames site.
