[MOBY-l] OWL-S, and RDF ontology for MOBY?

[Phillip Lord]

>>>>> "Frank" == Frank Gibbons <fgibbons at hms.harvard.edu> writes:

  Frank> Hey MOBYers,

  Frank> Couple of ideas for MOBY-S, wanted to throw them out there,
  Frank> see if they make any sense....

  Frank> 1. OWL-S is (as I understand it) an extension of OWL to web
  Frank>    services,
  Frank> enabling discovery and machine-machine negotiation of web
  Frank> services (that's the "-S" in OWL-S). Since OWL-S is RDF, and
  Frank> MOBY Central already produces RDF graphs as part of the
  Frank> registration process, it's really a question of making the
  Frank> RDF graphs OWL-S compatible. Is that something that's part of
  Frank> the plan for the ongoing re-write of MOBY-S?  (Why not)? I
  Frank> realise that OWL-S is evolving, so it might not be mature
  Frank> enough for use yet, but perhaps there are other reasons why
  Frank> it wouldn't be a good idea?

OWL-S is, as the name suggests, based on OWL rather than RDF. It is
certainly the case that you can represent OWL in RDF, but the full
semantics of OWL are quite a bit richer than those available in
RDF. So its not necessarily a trivial process to convert between
them. 

Early in the mygrid project we looked at what was then DAML-S for
incorporation into our service ontology...


@ARTICLE{mygrid-services,
  TITLE = {{A Suite of DAML+OIL Ontologies to Describe Bioinformatics Web  Services and Data}},
  AUTHOR = {Chris Wroe and Robert Stevens and  Carole Goble and Angus Roberts and Mark Greenwood},
  YEAR = {2003},
  JOURNAL = {the International Journal of Cooperative Information Systems},
  VOLUME = {12},
  NUMBER = {2},
  PAGES = {597--624},
  URL = {http://www.cs.man.ac.uk/~stevensr/papers/jcoopis-final.zip},
  KEY = {ontology,stevens}
}


There is also a recent paper that I co-authored with Mark and co,
which refers to some extent to OWL-S.  

@INPROCEEDINGS{iswc-bio-moby-comparison04,
  AUTHOR = {Phillip Lord and Sean Bechhofer and Mark
                  D. Wilkinson and Gary Schiltz and Damian Gessler and
                  Duncan Hull and Carole Goble and Lincoln Stein},
  TITLE = {Applying Semantic Web Services to bioinformatics:
                  Experiences gained, lessons learnt},
  BOOKTITLE = {International Semantic Web Conference},
  YEAR = 2004,
  NOTE = {Accepted For Publication},
  PDF = {download/publications/biomoby-comparison-iswc2004.pdf}
}

I think OWL-S is mostly aimed at automated composition which is not
really what moby-s is aimed at. You're unlikely to be able to get
semantic descriptions rich enough to describe services well enough to
be able to do this anyway, in which case you are accepting a lot of
complexity from OWL-S which will not necessarily buy you very much. 




  Frank> 2. I've been playing with extending the MOBY-S object
  Frank>    ontology, and
  Frank> talking with the BioPAX people about how they propose to
  Frank> integrate other ontologies into theirs, so that we might do
  Frank> the same. A month ago, at ISMB, it sounded like they were
  Frank> planning to import the PSI-MI ontology (for example)
  Frank> wholesale, and integrate it into theirs. Subsequent changes
  Frank> to PSI-MI would require manual inclusion. MOBY's ontology
  Frank> works in a similar way: in order to use PSI-MI, I've had to
  Frank> register each object separately. From this, and some reading
  Frank> I've been doing, it seems like the best way (and certainly
  Frank> the Semantic Way) to extend the ontology would be to have it
  Frank> be RDF-based, so that extension would involve simply
  Frank> including the URI of the ontology we wish to include. In this
  Frank> way, any subsequent changes in the included ontology are
  Frank> automatically included in MOBY's. How practical would it be
  Frank> to use this method in MOBY-S?  Presumably, that's what S-MOBY
  Frank> does - Gary?


Again there is an outstanding issue here. In the case of BioPAX they
are explicitly using OWL-DL, and its a bit of an unanswered question
as to what importing an ontology into another actually means. Most of
the tools can't cope with this, for example (some concepts would need
to be uneditable). On the flip side, the strong semantics of OWL-DL do
actually give you something to work on with ontology imports. It
should be possible to determine the relationship between concepts (so
long as they share at least some parental concepts in common) if they
are both defined in OWL-DL. 

With RDF, per se, its a little less clear. What happens, for example,
if two different imported ontologies defined the same concept in two
different ways? 

Cheers

Phil