From: Pete Haglich ([email protected])
Date: 03/10/03
I don't deny that the spatial knowledge is more useful if it is more specific, such as specific addresses or facilities in the examples given below. However, in our experience with our particular domain of discourse it is useful to speak of routes between cities or journeys in which the cities are treated at terminal points or way-points even though the "exact" location within the city is specified. For these purposes we consider cities to be idealized points. In the problem areas in which we work, we would say things like: <Road rdf:ID="Interstate80"> <hasTerminus rdf:about="#SanFrancisco"/> <hasTerminus rdf:about="NewYorkCity"/> </Road> <City rdf:ID="SanFrancisco"/> <City rdf:ID="NewYorkCity"/> We might then discuss whether or not the cities are connected based on the existence of serviceable routes between them. This discussion would parallel one based on: <Route rdf:ID="YahooRoute12slfjdsalfj132"> <hasTerminus rdf:about="Address234234sf21343"/> <hasTerminus rdf:about="Address230883094sdfz"/> </Route> <StreetAddress rdf:ID="Address 234234sf21343"/> <StreetAddress rdf:ID="Address230883094sdfz"/> in which we might consider routes connecting two addresses. I might point out that airports have 2 dimensional extent but are generally treated as points. It's all a matter of what grain size you want to adopt. For some of our applications we have found that a very coarse grain is appropriate. I guess that along with Austin I wanted to point out that a variable-grained spatial ontology has utility in some real world applications. On Monday, Mar 10, 2003, at 13:07 US/Eastern, Adam Pease wrote: >> As far as the nature of a city being a point vs being a region, I >> think it goes a bit deeper than what you suggest and it depends on >> the granularity of the discourse. For example, for travel >> itineraries, cities are often treated as points. > > Interesting. If I type a city into Yahoo and ask for directions, it > chooses a default address within the city. When I get a travel > itinerary, it's to a specific airport associated with a city. This > still seems like an interface issue. The objective reality is that a > city is a region. One may abstract it to a vague point in discourse > or GUI, but that doesn't change the facts of what a city is. If one > builds that abstraction into the ontology itself, that dramatically > limits one's ability to reuse the ontology in a different content, or > integrate with a GIS or other repositories of date, which are the > purported benefits of using ontologies to begin with. > > Adam
This archive was generated by hypermail 2.1.4 : 03/10/03 EST