On 5/10/07, Eric Scheid <[EMAIL PROTECTED]> wrote:
in XTM and others, the PSI isn't a sole attribute, but instead a given
category could have multiple PSI, and hopefully one or two in common
with the PSIs that you are using in your schema.
Eric is, of course, correct. An arbitrarily large, but hopefully small,
number of subject indicators can be associated with a single entry in an
ontology. In such a case, each of the subject indicators *should* be
considered by the creator of the category element to be semantically
equivelant. Thus, one might find the following:
<atom:category scheme="..bobs_ontology"
term="Communication.Publication.Book"
label="Book">
<atom:link rel="psi" href="http://de.wikipedia.org/wiki/Buch"/>
<atom:link rel="psi" href="http://en.wikipedia.org/wiki/Book"/>
</atom:category>
What that says is that the concept or thing which is categorized as "
Communication.Publication.Book" in bobs_ontology is referred to by humans as
"Book" and is the same thing which is identified by the listed German and
English Wikipedia URIs. If, in some other schema, (such as "Great
Inventions") someone else was to create an entry that linked to either of
the two Published Subject Indicator URIs, we would be able to infer
transitive similarity between the term for a specific "Great Invention" and
the term "Communication.Publication.Book" in bobs_ontology. In this way, of
course, we can support cross-ontology reasoning, translation between
languages, etc.
In a system that doesn't support PSIs or a similar concept, there is really
no way to reason about the meaning of any term unless you have intimate
knowledge of the schema that defines it. One might be able to establish
precise equivelances between terms in a small number of "popular"
ontologies, however, that is a very expensive process which requires a great
deal of communication overhead and continued maintenance as terms are added
or modified. It is better, I think, to let understanding emerge from an
analysis of the encoding of the data -- even though that understanding might
be a bit fuzzy.
bob wyman
