classification results, inferred object property hierarchy - visibility of entailments

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

classification results, inferred object property hierarchy - visibility of entailments

Alan Ruttenberg-2
There are a lot of uninformative entailments listed in this. Ideally they are filtered out, or at least a button is given in the view to filter them out. Leaving them in makes it difficult to find entailments of interest.

In classification results:

That properties are subPropertyOf topObjectProperty
Entailments about inverses of object properties (particularly when the inverse property is defined)

In the inferred object hierarchy.

That bottomObjectProperty is subPropertyOf every property.

--

In both the inferred class hierarchy and the inferred property hierarchies, it would be helpful to give a visual indication when the term is classified under a new parent. Since only red and black are currently used for label color, I suggest using blue to indicate a change in the place in the hierarchy.

In the object property view it isn't clear how one determines what elements are entailed, or which entailments are made.

For example, if I have p inverseOf p1, and p antisymmetric, then p1 is also antisymmetric. I don't see how we can find that out currently.

Thanks,
Alan

_______________________________________________
p4-feedback mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/p4-feedback
Reply | Threaded
Open this post in threaded view
|

Re: classification results, inferred object property hierarchy - visibility of entailments

Alan Ruttenberg-2
It is also interesting to see what happens if you ask an explanation of some property being inferred as being subproperty of topobjectproperty.

-Alan

On Sat, Jun 30, 2012 at 12:01 AM, Alan Ruttenberg <[hidden email]> wrote:
There are a lot of uninformative entailments listed in this. Ideally they are filtered out, or at least a button is given in the view to filter them out. Leaving them in makes it difficult to find entailments of interest.

In classification results:

That properties are subPropertyOf topObjectProperty
Entailments about inverses of object properties (particularly when the inverse property is defined)

In the inferred object hierarchy.

That bottomObjectProperty is subPropertyOf every property.

--

In both the inferred class hierarchy and the inferred property hierarchies, it would be helpful to give a visual indication when the term is classified under a new parent. Since only red and black are currently used for label color, I suggest using blue to indicate a change in the place in the hierarchy.

In the object property view it isn't clear how one determines what elements are entailed, or which entailments are made.

For example, if I have p inverseOf p1, and p antisymmetric, then p1 is also antisymmetric. I don't see how we can find that out currently.

Thanks,
Alan


_______________________________________________
p4-feedback mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/p4-feedback
Reply | Threaded
Open this post in threaded view
|

Re: classification results, inferred object property hierarchy - visibility of entailments

Timothy Redmond

There are a lot of uninformative entailments listed in this. Ideally they are filtered out, or at least a button is given in the view to filter them out. Leaving them in makes it difficult to find entailments of interest.

Interestingly Matthew and I were talking about this recently.  I have been filtering out trivial entailments as I see them for some time but there are still many left.  But I did this in a very non-systematic way so when Matthew asked what entailments are filtered out, there is no easy way to get an answer.  We decided that I should create a centralized new axiom visitor to do this in a systematic way.   By concentrating the logic in one place we will be able to easily say what gets filtered and why.  By using a visitor, it will be easier to get all the trivial entailments at once rather than incrementally adding filters over time.

I think that your idea of a button is probably a good idea.  A problem with filtering is that people will ask why certain entailments are missing. There is also the slippery slope problem.  It is easy to argue that entailments that don't depend on any asserted information should  be filtered.   It gets a little harder to see where to stop once you start filtering entailments that follow "too easily" from asserted information.

I don't think that the classification results view has any filtering at the moment.  It does display a lot of junk that makes it hard to use.  It also has some other problems - I believe that it doesn't refresh properly.  I am also unaware of filtering done on the inferred object hierarchy.

-Timothy



On 06/29/2012 09:05 PM, Alan Ruttenberg wrote:
It is also interesting to see what happens if you ask an explanation of some property being inferred as being subproperty of topobjectproperty.

-Alan

On Sat, Jun 30, 2012 at 12:01 AM, Alan Ruttenberg <[hidden email]> wrote:
There are a lot of uninformative entailments listed in this. Ideally they are filtered out, or at least a button is given in the view to filter them out. Leaving them in makes it difficult to find entailments of interest.

In classification results:

That properties are subPropertyOf topObjectProperty
Entailments about inverses of object properties (particularly when the inverse property is defined)

In the inferred object hierarchy.

That bottomObjectProperty is subPropertyOf every property.

--

In both the inferred class hierarchy and the inferred property hierarchies, it would be helpful to give a visual indication when the term is classified under a new parent. Since only red and black are currently used for label color, I suggest using blue to indicate a change in the place in the hierarchy.

In the object property view it isn't clear how one determines what elements are entailed, or which entailments are made.

For example, if I have p inverseOf p1, and p antisymmetric, then p1 is also antisymmetric. I don't see how we can find that out currently.

Thanks,
Alan



_______________________________________________
p4-feedback mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/p4-feedback



_______________________________________________
p4-feedback mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/p4-feedback