Skip to Content.

swikig - [swikig] Semantics of MW Categories and OWL

swikig AT

Subject: Semantic Wiki Interest Group

Text archives

[swikig] Semantics of MW Categories and OWL

Chronological Thread 
  • From: "Uschold, Michael F" <michael.f.uschold AT>
  • To: <mediawiki-l AT>, <semediawiki-user AT>, <swikig AT>
  • Cc: "Clark, Peter E" <clarkp AT>, "Kitzmiller, Ted" <ted.kitzmiller AT>, "Jones, David H" <david.h.jones AT>, "Folger, Deborah H" <deborah.h.folger AT>, "Murray, William R" <William.R.Murray AT>
  • Subject: [swikig] Semantics of MW Categories and OWL
  • Date: Fri, 3 Aug 2007 18:12:45 -0700
  • Envelope-to: swikig AT
  • List-archive: <>
  • List-id: Semantic Wiki Interest Group <>

The semantics of categories in Media Wiki seems to be too general, so that it is not amenable to linking up with OWL.
For example, by conflating subclass and instance with one relation, 'Category', I cannot capture the fact that say Person is an instance of OWLClass, and also a subclass of OWLThing. 
One specific problem arises when I create a Template for say Project.
I want every project created to be in Category Project, so I include "[[Category:Project]]" in the wiki page: "Template:Project". 
This gives the desired result that each wiki page using the template is in the project category.
However, it ALSO puts the template itself in the Project Category. This is fine if the semantics is: "is related to in some way". But I want the semantics of OWL: is a member of the classs. I don't know how to do this in SMW.
Another problem is that if I use another template on the same page (say Manager), then the page will be in both categories: Project and Manager.
Again, this is fine with the semantics being "is related to in some way".  But it gives the wrong answer if I want to interpret this as member/instance of a class.  A page that happed to use a macro that was about a certain topic is what I get. I want to clearly say that a wiki page is an instance of a class and not the broader meaning of "is related to in some way".
So the basic question is: how can I get the Semantic Media Wiki to distinguish different meanings of 'Catagory' that OWL supports and that I need given they seem to be all conflated in SMW.
I fear I will be doing a variety of workarounds. 
If anyone has any thoughts on this, I'd be grateful.
Here is my current understanding of a MW Category:
* Fundamentally, it seems to be a tag to associate with one or more wiki pages, much like a FlickR tag (e.g. Person, Africa)
* A Category can have sub-Categories (e.g. AirlinePilot for Person and South-Africa for Africa)
* A page that is associated with a sub-Category is also associated with all super-Categories (hierarchical inference of a sort)
* When a Category page is shown, you get to see all the pages in that category for free 
That seems to be about it. With such a weak semantics for Category, it can be used to represent many different things:
* subclass e.g. AirlinePilot is a subclass of Person
* instance e.g. Africa is an instance of a Country
                e.g. Joe is an instance of a SouthAfricaPerson
* subregion e.g. SouthAfrica is a subregion of Africa
* topic:     e.g. SouthAfricaBook is a book on the topic of SouthAfrica
As it is, this can give rise to undesirable inferences.
In OWL, we need to make clear distinctions between these kinds of relationships.
Suppose we have a category hierarchy as follows.
Suppose we have the following wiki pages:
Wiki Page    Description                      Category
Joe          A person living in South Africa  SouthAfricanPerson
Apartheid    A book about South Africa        SouthAfricaBook
MediaWiki will infer:
That JOE:  is in the following categories:
* SouthAfrica, Africa, Country, Person
That APARTHEID is in the following categories:
Book, PhysicalObject, SouthAfrica, Africa, Country
Are these desirable inferences?
Q: What is the relationship between the book Apartheid and the category Country?
A: It is a book about a place that happens to be in a particular country.
Well, virtually all places are in some country, so such inferences will often be irrelevant and unwanted.
IHMO, the root of the problem is that the semantics of Category is very weak - it can be used to represent any arbitrary association*
* I do recognize a tradeoff here: a weak semantics makes it easier for the masses to do enter information.

Michael Uschold
M&CT, Phantom Works
425 373-2845
michael.f.uschold AT 

COOL TIP: to skip the phone menu tree and get a human on the phone, go to:


Archive powered by MHonArc 2.6.16.

Top of Page