[MMSEM-UC] Comments on Algorithm Representation UC

Dear all,

as promised in the last telecon, please find some comments on the  
algorithm representation use case [1] in this email.

As Raphaël pointed out in his review [2], I think this use case fits  
very well in the work of the group as it tries to highlight  
interoperability issues in multimedia processing and how processes can  
be realized/improved through having semantic descriptions of the  
content that is processed.

I have some comments on your use case and its description in the WIKI:

First of all, you only explained an example including singular  
processing steps, what perhaps would highlight the problems more would  
be an example including more than one steps to really see where  
interoperability problems can occur.

Perhaps you could also formalize the interoperability issues more as I  
see them in four different flavors:
1) A <-> A (between algorithms)
2) A <-> D (between algorithms and data)
3) D <-> D (between data and data)
4) O <-> 0 (between ontologies used to describe content and algorithms)

An interesting questions would perhaps be: where can data interop.  
occur (I see this in matching input/output of algorithms)?
How can you overcome interoperability issues?
Where do you need detailed semantic descriptions of content?
How can these exactly be used to improve processes?

What I think would help your use case, would be to provide a running  
example / scenario (this could be your radiology usecase) ->
what different algorithms could be used in this case? How should  
input/output look like - what are the most important properties that  
it should have (on the metadata level)? What state do you need to  
record during processing? ...

Your use case reminds me very much of the problems that the Semantic  
Web Service (SWS) - Research aims to tackle: In the WSMO [3] approach  
for example you have
1) Formal descriptions of executable functions (Web Services)-> this  
is similar to your formal description in the algorithm ontology
2) Capabilities (what can you achieve by executing this function) ->  
this is similar to the goal of the algorithm in your case
3) Input / Output
4) ontologies describing the functionality
5) mediators allowing to overcome interoperability issues between  
input/output formats, goals or ontologies

Perhaps you might have a look at [4].

You might also have a look at the GRISINO project [5]. In this project  
we aim to realise similar things:
We try to match content properties with SWS functionality to enable  
(semi-)automatic distributed multimedia processing.

I hope some comments help...

Best regards,

Tobias


[1]  
http://www.w3.org/2005/Incubator/mmsem/wiki/Algorithm_representation_Use_case
[2] http://lists.w3.org/Archives/Public/public-xg-mmsem/2006Oct/0061.html
[3] http://www.w3.org/Submission/WSMO-primer/
[4] http://www.w3.org/2002/ws/swsig/
[5] http://www.grisino.at

-- 
________________________________________
Tobias Bürger, http://www.deri.org
skype: tobitrautich

Received on Tuesday, 21 November 2006 06:04:42 UTC