| | Today we will discuss open-world reasoning and the difficulties it poses to our ontology design.
Recall our previous implementation of the N-ary ODP. It included class definitions for SequenceItem and FinalSequenceItem. The two classes are identical except FinalSequenceItem is restricted to individuals who do not have any additional items following them (a maxCardinality of zero on the NextSequence relationship). This means that any SequenceItem without a nextSequence relationship is also a FinalSequenceItem.
Let's define a SequenceItem called testItem with no NextSequence relationship. We might expect our reasoner to classify our testItem as a FinalSequenceItem, since we have not specified a NextSequence relationship. This, however, is not the case. OWL was designed using the open world assumption. In systems utilizing the OWA, facts that are not stated are considered unknown, rather than false. In the testItem example, an open world system can only say that the testItem has at least zero NextSequence relationships. It can not say that it has exactly zero NextSequence relationships, because we have not explicitly stated as much (recall: the omission of a fact does not imply its nonexistence).
This issue comes up again in our definition of CharacterSequenceItem. If we declare an individual, testCharacterItem, from the class SequenceItem with a property value of the Character "K" for the ContainsData relationship, we can not infer that our testCharacterItem is a CharacterSequenceItem. The testCharacterItem could contain other, non-specified, non-character data. If, however, we specify a cardinality-constraint of one for the ContainsData property of a SequenceItem, a reasoner can now infer that our testCharacterItem is, in fact, a CharacterSequenceItem. Since each SequenceItem can have at most one ContainsData property value, we have explicitly constructed a closed-world environment around the ContainsData property.
(More on this tomorrow... this post needs edited too... it's not very easy to read. Does anyone understand anything I've said here today? How can I make it more clear?)
|
| | Posted 9/5/2006 4:46 PM - 61 Views - 0 eProps - 0 comments
- recommend
  - recs0
- share
- email
 - sent0
|
