[MOBY-dev] BioMOBY and myGrid: my $0.02

[Duncan Hull]

Hi Mark

Mark Wilkinson wrote:
> Thanks for the offer!  If you could provide that list I will 
> incorporate it into the text and add you as an "author" :-)

Well you've probably thought of all these already, but in case you 
missed any, here they are:

(Your mileage may vary of course...)

Pros for tagging, BioMOBY style:
========================
+ Low barrier to adoption
+ Allows many people to annotate services
+ Wisdom of crowds, potentially captures more consensual knowledge


Cons for tagging, BioMOBY style
========================
+ The wisdom of crowds is always "close to madness" [1], you can cite 
Alcuin on that one!
+ Can produce a large quantity of low quality metadata, what the Flickr 
people call "noisy metadata" [2]
+ Difficult to precisely and accurately query registry


Pros for curation, myGrid style
======================
+ Less noise in service metadata
+ Potentially more coherent unified view
+ Higher quality (?), lower coverage


Cons for curation, myGrid style
======================
- Services not annotated by default, as with BioMOBY
- Restricts the number of people who can annotate services
- Labour intensive
- Time consuming
- Expensive
- Captures a smaller world view, the "wisdom of the curator" rather than 
the "wisdom of the service provider"


> (actually, NOT having you as an author was a bit of an embarrassing 
> oversight anyway, given that you have now done two graduate degrees 
> studying meta-Moby!)
The third degree isn't in the bag just yet ;)


While we're at it...

Pros for DL reasoning during service discovery
=================================
+ More intelligent matchmaking e.g.
+ Better precision
+ Better recall
+ Ability to rank services which are otherwise unrankable (without 
invading users privacy).
+ Rank by degree of match between request and advertisement [3])
+ Analagous to an "Advanced Search" 
http://www.google.co.uk/advanced_search?hl=en

Cons for DL reasoning during service discovery
=================================
- Queries have to be written either in SPARQL or as OWL class expressions
- Few people used "Advanced Search" features
- Requires detailed annotation of services using OWL
- Issues with scaling

If any of this doesn't make sense, let me know and I can clarify.

Duncan


[1] http://www.cs.man.ac.uk/~hulld/q2004-02-11.html
[2] http://dx.doi.org/10.1145/1125451.1125462
[3] http://eprints.ecs.soton.ac.uk/7606/

-- 
Duncan Hull
http://www.cs.man.ac.uk/~hulld/
+44 (0) 161 306 5139