[topicmapmail] Proceedings of Semantic Technologies for eGov

[Murray Altheim]

rhodgson wrote:
[...]
> It is clear that it is easy to incite people in your community - why? It
> seems that we provoked these reactions because of one statement we made
> that "we do not recommend topicmaps for semantic integration". Meaning
> data integration that requires rules and inferencing for data integrity
> and consolidation. We know of no cases where topicmaps are used for
> these applications. In your opinion are there any areas of semantic
> technologies where topicmaps are not appropriate?
> 
> You should know that we do recommend topicmaps for other needs. Such as
> content management, intelligent search, document indexing. All cases
> where "information objects" are more at work than "business objects".

Ralph,

I see two important parts to the Topic Map model. One is the internal
graph structure of nodes and edges, or in Topic Map parlance, Topics
and Associations. Where RDF has tuples or triples, Topic Maps has the
concept of Scope, which is the context in which an Association is made.
This particular graph model is essentially all one needs to do
interesting "knowledge modelling", i.e., anything that can be done in
RDF can be done in the "internal" graph of Topic Maps. I spent a fair
amount of time writing a translation from the Cyc ontology way to XTM
back in 2000/2001, and while I didn't finish the work (due to various
factors both mine and Cyc's), I see no technical barrier to this
working, nor to a Topic Map-based inferencing system. All that one
would need is a set of XTM documents for the logical foundation,
which could be a translation from Protege or any other existing
language. It's just a graph. I can imagine that if I had the funding
I could build such an engine on top of Kal Ahmed's TM4J.

The second part of Topic Maps is the part where it excels over RDF-
based applications, since it is *mapping* things that Topic Maps was
really designed to accomplish. Whereas the RDF community is still
working over the meaning of reification and URIs, etc., you can
reliably map addressable and non-addressable subjects right now in
Topic Maps. This takes the basic idea of an "ontology" described above
and maps from it to a client's information resources (addressable or
non-addressable).

Now, in almost any text on Topic Maps you'll find talk about
integration of information resources, using scope, merging and
other Topic Map features (search on "topic maps" and "infoglut"). The
combination of a Topic Map-based inferencing engine, some domain-based
ontologies, and the ability to map information resources, well, that
seems like a pretty powerful combination to me. I'm writing my Ph.D.
on everything described above except the inferencing engine (since my
work is focused on informal ontologies).

You might find that the few Topic Map vendors don't talk extensively
about private contracts they have either already completed or are
currently in progress. It's a small community and not everyone talks
about the details of their projects publicly -- I know of several
vendors/contractors' work and they don't talk about it publicly,
period. That doesn't mean that Topic Maps aren't appropriate for
"semantic integration" applications, nor that they aren't already
being used in that capacity. I see no technical barriers; perhaps
social/community/political.

Murray

......................................................................
Murray Altheim                    http://kmi.open.ac.uk/people/murray/
Knowledge Media Institute
The Open University, Milton Keynes, Bucks, MK7 6AA, UK               .

   The world, Bush said, is now riven by "the clearest of divides:
   between those who seek order and those who spread chaos; between
   those who work for peaceful change and those who adopt the methods
   of gangsters; between those who honor the rights of man and those
   who deliberately take the lives of men and women and children
   without mercy or shame. Between these alternatives there is no
   neutral ground."
   http://www.sfgate.com/cgi-bin/article.cgi?file=/c/a/2003/09/24/MN298975.DTL

   Actions speak louder than words, and I think it's pretty clear
   which side of the divide Bush is on.