We've completed initial analysis of generic patterns behind IIP TIP set. TIP Manager ( http://iringsandbox.org:8080/tip/tipmanager ) database was received from IIP team (as of Nov 14th) and processed in the following way:

- Total 250 TIPs with mapping status "Completed" were selected.

- Mappings for selected templates were transformed in the .15926 Pattern format (specs for our open pattern libraries downloadable from download/file.php?id=148).

- Total 109 TIPs were identified with two or more templates in the mapping.

- All non-trivial mappings were stripped of role restrictions by particular classes and compared together to identify 30 generic patterns.

- Generic patterns were named to describe the nature of relationships between top-level Possessor parameter and value parameter(s) (remember that all TIPs so far are designed as object-property informayion models).

For example, analysis of two TIPS:

Fluid Category
Fluid Code

reveals the same three templates with roles restricted by varying classes:

AssemblyOfIndividualTemplate with characterised object (Possessor) playing hasWhole role .
ClassifiedContainmentOfIndividual with hasContainer role occupied by the same entity as hasPart role of the previous template.
ClassificationOfIndividual with hasIndividual role occupied by the same entity as hasContained of the previous template.

Therefore the generic pattern behind these two TIPs was named ClassificationOfClasifiedContentOfPartInAssembly.

The list of all 30 generic patterns with names of TIPs based on the corresponding pattern can be downloaded from
Generic patterns in .15926 format (Python dictionaries) can be downloaded from
Use this file to view generic pattern structures (in any text editor). Loading them into .15926 Editor is possible, but of a little business use with the current version. Full library of TIPs for use in .15926 environment will be published with next version of the Editor which will have enhanced pattern management and mapping capabilities.

What I miss is a definition of the semantics of these patterns.
The names, like IndirectPropertyOfClassifiedContentOfPartInAssembly, are meaningless to me.

Hans,

I've described in detail the structure of ClassificationOfClasifiedContentOfPartInAssembly pattern above.



Is such description meaningful?

I am afraid not. Quite frankly I have no idea what
is supposed to mean in engineering terms.

Let me try:
Object#1 has object#2 as a part in an assembly;
Object#2 contains object#3;
Object#3 is classified.

or in a narrative format something like: Classification of the content of a PossibleIndividual that is a part of an assembled PossibleIndividual.
(no doubt that there are no temporal parts involved, so the content will always remain in place)

So for example:
Boiler X has Drum Y as a part, and Drum Y contains Fluid Z that is classified as BoilerFeedWater.

Great! But what does this buy you?

And if you have a not-controllable desire for this, then call it something like ClassificationOfContentOfPartInAssembly

The number of permutations of this denormalization exercise will be mind-boggling.
Besides that, if anything in that combination changes, you have to redefine everything. And I shudder when I think of the wild SPARQL queries that will often give no outcome because some bloke found yet another nice shortcut. Yuk!!!
I can't stop you guys, but one day you will deplore having gone this route.

Great indeed! You have reworded it, but it's 100% what I've tried to say.



This is a wrong question.

As well as I understand the IIP process, they've just discovered that in some legacy system they have FliudCategory and FluidCode recorded as a property of a Boiler. What was the right way to proceed from this point in your opinion (leaving aside for a moment temporal aspects of the modelling)?

If you want to create a pattern for each and every ill-designed property then you are in real trouble. Why should we standardize that? There will be no end to that!

What is wrong with letting each software supplier think about what the explicit semantics of his data are, and then select one, two, or more templates to model that for his mapping? Perhaps this is something for Part 11?

Sorry, Hans, your position on that is beyond my comprehension. I'm living in the world where software suppliers (except esteemed Bentley, of course ) pay little to no attention to semantic adaptors for their data.

But if the best of software suppliers discovers such a situation in his product - his only option will be to choose the same 3 templates in the same order, no? That's why it is a generic pattern repeated several times for the real properties of real software.

What are your criteria to choose models worthy of a "template" name? Some of your graphs are no less complex then the one for ClassificationOfContentOfPartInAssembly.

The least they can do is to come up with a narrative describing the semantics in engineering terms.
Just try four of five.

My question also - is there a system in their approach? I hope my generic pattern analysis will help them to formulate WHAT models they do.

Your pattern ClassificationOfContentOfPartInAssembly translates to the following graph:



Does your pattern have roles pointing at the yellow and the green instances of PossibleIndividual?
If not the compound relationship between the boiler system and the class of the contained fluid can only be reconstructed via lifted templates.
If so, what do you gain with this denormalization? Can you fathom the combinatory explosion?
And how do you know which of that myriad of patterns to use in your SPARQL query when you want to know what the part in the assembly is?