Minutes Meeting 22 September + QCR debate

Hi,

The minutes from today's meeting are online:
http://www.w3.org/2010/09/22-ssn-minutes.html 

Summary: 
- Comments on Raul's figures (Ontology Structure) and documentation of the DUL alignment 
- Progress and comments on SSN Observation page and kitchen example (Andriy) 
- Discussion on the use of QCRs in the ontology (Qualified Cardinality Restriction)

Plus: my ideas (not expressed at the meeting) on what to do next
- 


ONTOLOGY STRUCTURE ( --> Raul): 
http://www.w3.org/2005/Incubator/ssn/wiki/Report_Ontology_Structure 

- Figure 1 is judged too complex for newcomers to the ontology 

The suggestion is to have a figure like the one at the top of this page to introduce the module (ideally using the same layout as Figure 1)

- but Figure 1 as it stands now and Figure 3 should be used later in the text because it is valuable as an all in one picture (ant out the legend either before them or at the bottom of the page but not between them) 

- Figure 4 (the DUL alignment) triggered many comments because it helps to clarify the alignment for us who knows that we’ve done. What we can do to help "new readers", to better visualise what's from DUL from what's from the SSN XG, is to add a "dul:" prefix to the name, put a larger space between each parts of the figure and draw the alignment subclass links differently - maybe a different colour? in a way which suggest to the user that they "are" the alignment. Also, the DUL properties should be removed to further simplify the picture.

DOCUMENTATION OF THE DUL ALIGNMENT (-->Krzysztof and Michael)

Krzysztof and Michael need to deliver the DUL page as soon as possible because Raul needs to tune his page against it. 

They intend to adapt the content of their paper submitted to SSN10 
http://www.personal.psu.edu/kuj13/SSO%202010.pdf 

SSN OBSERVATION 

Example reworked by Andriy:  http://www.w3.org/2005/Incubator/ssn/wiki/SSN_Observation 

Also need to check consistency of the content of this page with last version of the ontology and with the documentation of the DUL alignment (Laurent to do another pass on the figures)

QUALIFIED CARDINALITY RESTRICTIONS OKAY VOTE (--> complement to minutes)

Context: 
- late addition by Michael of QCR for the Observation class (e.g. for the , which has not been reviewed by the group, inspired by similar modelling decisions made in O&M

Discussion: 
- specific risks flagged by Laurent: lack of support by DL reasoners in case of errors, possible mismatch between the modelling intent in O&M (closed world) and the SSN XG scope (open world), impact on closure date of XG (hence the suggestion to postpone)
- argument by Michael that we should not restrict our modelling because of weaknesses in error reporting in tools and suggestion by Krzysztof that DUL is evolving in the same direction,
- Raised awareness that we may have engaged in heterogeneous modelling practice: our work on the ontology prior and after the DUL alignment and now the addition of QCR for the O&M-inspired part: Cory noting that it is important to make sure that there is no part that is way more complex than other parts.

Vote: no clear majority  
- we voted on Go/Postpone/NoGo (on use of QCR in general not on each individual modelling decisions) without a clear majority (1 no go, 2 postpone (including Raul who was not present), 1 maybe postpone, 3 go, 1 go or postpone (later becoming a go). 

Decision: okay for having a go at using QCRs but with a more detailed analysis of the consequences of the modelling decisions (we have 4 QCRs to review - see below), and enough documentation to back them, and if required the identification of the possible counter-measures for the confirmed risks (advice or announce of future work for the extension of the model for non "QCR=1" use cases: compound observations, compound properties, compound sensors, compound features). All this in a very short time, so everyone who voted yes is expected to contribute strongly NOW!


MODELING ISSUES AND QCR (--> please contribute as soon as possible!) 

QCR 101: 4 for a car has 4 wheels, or a Four Cheese Pizza has 4 cheese toppings

SSN has 4 QCR in total, all attached to the QCR classes:
- QCR 1 for sensingMethodUsed property and Sensing class
- QCR 1 for featureOfInterest property and FeatureOfInterest class
- QCR 1 for observedBy property and Sensor class
- QCR 1 for observedProperty and Property class

DUL (checked 2.4) does not use QCR!
Comment for Transition class: "A full representation of the transition ontology is outside the expressivity of OWL, because we would need qualified cardinality restrictions, coreference, property equivalence, and property composition." 

O&M and QCR=1 cardinality for the observation class 

I'm told by colleagues that this is a not a done deal at OGC. The use case for "Composite Phenomenon" is important! 

See Dominic Lowe, SWE DWG: The future of the SWE Phenomenon/Property model 74th OGC TC Toulouse, France

Use Case 2: Multiple phenomena
● Referencing multiple definitions:
● e.g. Satellite measuring 'methane, ozone, CO2'
● Something like SWE Composite Phenomenon needed.


My 2 cents: the part of the O&M model with cardinality 1 restriction DOES NOT COVER the "compound objects" - the assumption is that they will be managed outside the model. If we maintain the QCR-1 in our ontology, we need to announce / plan for extra classes and properties supporting these more complex use cases.


O&M vs. SSN and SDUL vs. SSN - CALL FOR YOUR INPUTS (--> please contribute as soon as possible!)

For me, today's vote reveals mainly that the issue of how our ontology is positioned against O&M is not settled and also that we have not fully anticipated the consequences of aligning it with DUL. Maybe we should add a part in the report discussing these issues near the conclusion. I welcome your comments/inputs on this matter.

Of my two concerns this morning, I think that I can probably live with the lack of debugging support (a possible solution is to build specific validation tools to detect the sources of errors in a different environment). But I maintain that we will need sooner or later to handle the "compound objects" use cases and not just the "single object" ones. 

Cheers
Laurent

PS: Thanks to Raul for sending the new figure of the ontology and to Andriy for extending his example to cover the Observation ontology (I'll keep you posted).

Received on Thursday, 23 September 2010 13:34:15 UTC