The use of the term "ambiguity" is often confusing when employed by
different communities. What Mike is getting at I think is that,
fundamentally, unless one enumerates all the members of a set, or unless
the set is an abstract entity such as an equilateral triangle, there will
be ambiguity in the set definition. For any definition, there will always
be "boundary cases" which are not clearly in or out of the set.
Suppose one defines a tank as an armored, tracked vehicle, then is an
APC (armored personnel carrier) a tank? Then maybe one defines the purpose
of the vehicle as delivering fire rather than troop transport, but what
about if the vehicle is then used in a pinch to move a small number of
troops? One might narrow the definition to address only intended rather
than actual use but this can go on indefinitely. In practice, it is rarely
possible to enumerate all the boundary cases. Human common sense is
brought to bear. But if we expect software to perform the classification
or reasoning task, human judgement is not available during the
computational loop. Of course, this issue is not necessarily limited to
automated computation. Even the best directions or definitions can have
loopholes for the human interpreter.
Languages such as SQL or UML offer some ability to represent terms that
can name the definitions as discussed above, but fall short of the
expressiveness of human language, and therefore often employ natural
language documentation to make clear what the formal language leaves
implicit. The advantage of formal languages such as first order logic are
that they offer an expressiveness more comparable to human languages, at
the price of low general familiarity to the software engineering community. (01)
At 08:31 PM 10/29/2002 -0500, MDaconta@xxxxxxx wrote:
>In a message dated 10/29/2002 4:57:56 PM US Mountain Standard Time,
>>You are creating an IC ontology. What is IC?
>Sorry about the acronym without explanation. IC stands for
>>Can you give an example of some descriptions from your ontology that
>>describe[s] the subjects of resources in the knowledge base“.
>The ontology will have concepts from each of the Intelligence domains.
>Some examples are Facility, Person, WeaponSystem, etc.
>Data sources will be explicitly mapped to these classes via a registration
>>Most of the time, there is no such thing as an unambiguous description,
>>exceptions arise for mathematical creations. The idea of an ontology is
>>to reduce ambiguity as much as is necessary for a given application. NB,
>>I did not say as much as POSSIBLE. That will often be a waste of time.
>The whole concept of URIs replacing words is to allow unambiguous description.
>For example, given a Department of Defense URI for a Tank like
><www.vkb.htm#Tank>www.vkb.org/Army/Equipment/#Tank I can unambigously
>state that I am
>referring to a weapon system and not a water storage device.
>Michael C. Daconta
>Director, Web & Technology Services
To post messages mailto:ontolog@xxxxxxxxxxxxxxx
An archive of the [ontolog] forum can be found
at http://ontolog.cim3.org/forums/ontolog (04)