4.4 The Web Ontology Language 85
Reasoning in an open world is sometimes counterintuitive. As an example
of this, suppose that we make the additional requirement that every gene
belong to exactly one species. This can be specified by adding the following
to the specification above:
<FunctionalProperty rdf:about="#occursIn"/>If one has not specified that thehbae3gene occurs in any species, then one
would infer that there is exactly one, as yet unknown, species where this
gene occurs. This is shown in figure 4.7.
Figure 4.7 An example of an unspecified but mandatory relationship. The blank
node represents the anonymous species that is necessary to fulfill the requirement.
The inferred resource and relationship are shown in gray.
Now suppose that one specifies that thehbae3gene occurs in two species:
<Gene rdf:about="hbae3">
<occursIn rdf:resource="#D.rerio"/>
<occursIn rdf:resource="#D.danglia"/>
</Gene>In a closed world, this would be inconsistent because the ontology allows a
gene to belong to only one species. In an open world, one would infer thatD.
reriois the same species asD. danglia. This is shown graphically in figure 4.8.
The inference thatD. rerioandD. dangliaare the same species is clearly
incorrect. There are many ways to remedy this erroneous conclusion. One
could explicitly specify that the two species are different as follows:
<Gene rdf:about="#D.rerio">
<owl:differentFrom rdf:about="#D.danglia"/>
</Gene>This would cause an OWL processor to signal an inconsistency exactly as in a
closed world system. One could also drop the requirement that genes occur
in only one species, which is more realistic. Doing this would eliminate the
spurious inference in this case, but it would not prevent the inference from