WS2 Working Notes 17 November

Overview

Discussed the remaining issues for the technical aspects of the standard. These will be changes to be made to the final version of the RFC.

Detailed Notes

Named Graphs

We will not use this approach in this release.

Facts

Question: How did we get over the fact that dependency on relationships can't be shown in EA.

Action: Need to talk to Sam Mancarella at Sparx about a possible workaround.

Example:

Have a tag on the stereotype that gives the name of the Annotation Property.

Fact = tagged value in this case.

So we use Dependency for class to Class relationships.

Where we require a dependency on a dependency is to link the Fact to the Annotation Property. Rather than the second dependency to link to the predicate we would need the name of the annotation property.

Notes on New EA File 'ODM MGB Edits'

This is a file in which MB has taken the EA version of the ODM Profile, as supplied by Sam Mancarella of Sparx Systems, and begun working through this to identify what the new RFC standardized set of ODM terms should be. There appeared to be some bugs, which need looking into, but some of these questions may simply be resolved by going through what the ODM profile should look like.

Note that as well as diagrams for the main features of RDFS and OWL, there are additional '99' diagrams containing a number of constructs which are present in the MagicDraw ODM material but which may not form part of the formal ODM standard (correct me if I've represented that incorrectly).

Fact

We went here to look for the term 'fact', which is intended the instance of an OWL Annotation Property, and would therefore be used for instances of DC 'Source' and terms derived from this such as 'Term Origin' (the 'termOrigin' metaterm).

We found 'fact' in the diagrams '7. Graphs and Statements' and in '99. Experimental'. The following notes were made:

Diagram: 7. Graphs and Statements

Diagram Note: Curious that this is not in the metamodel, showing a relationship between the Fact and that which it is a fact of (an instance of) i.e. the Annotation Property.

Discussion: In fact this is present. 'fact' is a sub class of 'triple' and this has UML Attributes for subject, predicate and object, with 'subject' and 'object' having the type of 'node'. Therefore the relationship between the fact and the subject and object of that fact are achieved by this type relationship in the UML Attribute.

'subject', 'predicate' and 'object' also exist in the ODM model as stereotypes, which each extend 'dependency'.

Diagram Note: In ODM Fact may be either an instance specification or a dependency.

For the purposes of EA, if we used the IS instead of Dependency to represent the annotation, then we should be able to use the Dependency to point at the Predicate.

Action: try both of these things.

Discussion: We had been focusing on 'fact' as an instance of OWL Annotation Property since this is how we intend to use it, however as defined in general, fact may be an instance of any kind of property. For this reason, there are a couple of available approaches to modeling these.

This needs to be understood in more detail, and will be the subject of a future separate session with the modeling experts.

Note attached to 'Slot' ('fact' extends 'Slot'):

Not in ODM

Discussion: the use of the base class 'Slot' for the term 'fact' is not part of the ODM standard.

Note that at present in these diagrams there is no clear delineation of what terms are or are not in ODM. These diagrams were extended by MB in this model file to make all the base classes and extension relationships visible, since none were visible in the received file.

In this case we looked for the base class of 'Slot' in EA and it is not made available (it is possible to name base classes to anything one likes but there is a selection menu of actual UML base classes in the EA editor). Slot is presumably not a known UML base class (the concept comes from frame based ontology modeling I believe).

Diagram: 99. Experimental

Diagram Note: There is a possible issue here:

Given that Fact is a subclass of Triple and that already has subject predicate object, while we are also considering having these as dependencies. However, that latter may or may not be simply a working arrangement for getting stuff out of the existing EA model content.

There may or may not be issues with the actual ODM Profile itself on this - to be dealt with separately.

Discussion: As noted above, there are 'subject' 'predicate' and 'object' as attributes of 'triple' and again as stereotypes in their own right. Here we are trying to tease out a little more how these two sets of terms relate to each other and how to apply this in the FIBO modeling, as well as how to determine a sub-set of ODM to publish as the FIBO implementation of ODM.

Action: Separate meeting to be held between Elisa, Pete and Mike to chase down these issues.

Classification Facets:

Action: MB and DC to talk about detailed terms to use Present proposals to EK/PR or the wider group on how to metamodel these.

AOB

URIs still to be dealt with.

 
ws2_thursday_17_november.txt · Last modified: 2011/11/28 15:00 by mikehypercube
 
OMG Home Logos and Trademarks Become a Member Become a Sponsor Upcoming TC Meeting TOP