[MOBY-dev] summary of research into semantics

[Phillip Lord]

>>>>> "Andrew" == Andrew D Farmer <adf at ncgr.org> writes:

  Andrew> (Trying again, last post got bounced for being too long...)

  Andrew> Hi all- I'm going to post a rough summary of my exploration
  Andrew> of various major efforts for defining standards for
  Andrew> representing "semantics". It will hopefully be of some help
  Andrew> in communicating the general trends in this area (to those
  Andrew> of us who aren't already intimately familiar with the
  Andrew> history of the field), and may help to provide some
  Andrew> grounding for future discussions of MOBY's requirements in
  Andrew> this area.

  Andrew> In particular, I think it raises some questions with respect
  Andrew> to the role of XML Schema in the current MOBY system (I'd be
  Andrew> interested to hear your thoughts on this Mark, especially in
  Andrew> light of your recent announcement of your intention to
  Andrew> revamp some aspects of this in the current system). Also,
  Andrew> I'd be interested in hearing other people's opinions of the
  Andrew> applicability of the RDF model to our needs.

  Andrew> Apologies for any errors or significant omissions. Tear
  Andrew> apart as you see fit...


My own feeling is that it is not at all clear yet, which technology is
going to be best for describing the semantics of services within
bioinformatics. At Manchester we have moved forward with OWL (or
rather DAML+OIL) representations of service metadata. This is very
rich, and expressive, but also fairly complex. Currently biomoby
uses a much simpler representation, which is much easier to
understand, but perhaps less adaptable. 

Good bad? Well I don't know. Using a simple representation is a pain
in the ass, if you need to represent complex ideas. In other words
simple, becomes simple minded. There again I'm sure we have all had to
wade through some maniac coders library which is so complex its not
worth learning. 

My idea is to try and make the semantic aspects of mygrid as
"pluggable" as possible, so that we can try describing services using
schmema, RDF, or OWL, or whatever, and see how much complexity we
need. My hope is that, if I can provide, for instance, a semantic
discovery service, generically enough, it should be possible to hook
it up to biomoby as well. I've wanted for a long time to take
biomoby's type system, and plug it into our technology, but sadly have
never had the time. 

Anyway I'm sorry this is a somewhat tangential reply to your
remarks. 

Cheers

Phil