Is a

Discussion (somewhat historical)
Effects of having "is a"
 * (+) Simpler Syntax because all Properties are in the Wikipedia Namespace.
 * (-) All Elements of XML Schema, RDF, RDFS, OWL, OWLS, and some others, would need to be Part of MediaWiki. This blows up the Wikipedia Namespace.
 * (-) If people specify directly whether they want to use RDFS or OWL or whatever, then Wikipedia becomes dependent on these particular technologies. It would be preferrable if relations correspond to a fixed intuitive meaning within the wiki, and are later mapped to whatever technical formalism is considered appropriate for applications. Of course, it must be clear what "is a" means. But the according "instance of" from RDFS and OWL are still quite incompatible in many sitiuation (having a totally different formal semantics), and it would be a pity if we end up in a two-worlds scenario of people using OWL and people using RDFS. For "Category:", both RDFS and OWL would be applicable without violating the intuitive semantics, and everybody who uses Wikipedia's data is free to choose which language to use.

What actually happens with MediaWiki categories in SMW...
This keeps coming up on the semediawiki- mailing lists. Semantic query treats category membership as transitive, so a query on category:Cities finds articles in its subcategories. And help on RDF Export says how SMW exports category properties (as rdf:type and rdfs:subClassOf in SMW 0.7). -- Skierpage 23:26, 10 August 2007 (CEST)

... Back to the discussion
If "yes", then this leaves me uneasy. But as a relatively new user of Wikipedia and SematicMediaWiki, I'm hesitant to register a strong opinion.

I would think that Wikipedia Categories would have fairly general and generic meanings, so the Category system would be both stable and consistent. To me, "Category" implies a subject taxonomy, although I know it's used for many other article classes(e.g. Orphaned Pages, etc.)

If the "is a" relation applies to an entire article, I can think of several examples where a separate "is a" relation would be very useful. Most important for my purposes would be to assign articles to various levels of discussion, especially when one article is meta-level discussion about other articles.

Moreover, "Category" seems most useful for navigation purposes, and the Category system is fully visible. In contrast, the "is a" relation could have many more uses, especially for use by automatic processing where the full complexity of the "is a" relations" would be hidden. One example might be layout heuristics for topic maps, where the "is a" relation is used to guide placement of nodes in the map.

Again, I may be wrong about all this. I just know that "is a" is the most common relation in semantic networks. If it's fully subsumed by "Category" then fine. --Russell Thomas 10:26, 9 December 2005 (CET)


 * The Categories are parially organized like "is a", but there are not the same. "is a" means that something is a specialization of something other. A Category is a group of things with similar meanings. So Categories are even much like Relation:context of or even just like "has something to do with" without decleared nearer. Categories are very subjective colored, but "is a" results even in an locial meaning.
 * MovGP0 01:33, 10 December 2005 (CET)


 * Great answer. Thanks.  Like I said, my gut reaction is that there is probably a need for "is a" relation.  I'm going to put up some sample pages from my world to play with various relations and attributes.--Russell Thomas 02:54, 10 December 2005 (CET)

Does including all elements of OWL blowup Wikipedia Namespace?
>> (-) All Elements of XML Schema, RDF, RDFS, OWL, OWLS, and some others, would need to be Part of MediaWiki. This blows up the Wikipedia Namespace. <<

I don't understand this statement. I would think that incorporating all or most of OWL would be a great thing, since that's where the Semantic Web community is going. Intuitively, I would have expected SemanticMediaWiki to fully embrace OWL (at least OWL lite). Am I missing something?? --Russell Thomas 10:36, 9 December 2005 (CET)


 * I think I can answer this Question from two viewpoints:
 * Implementation: The Example of WikiOnt shows, that some elements might be a good thing, but can get replaced by others or even might be useless in practice. OWL-Support also don't seems to be a requirement for the 1.x Versions of SemaWiki. More important, it needs a well working Wiki-Syntax to RDF-Mapping, and Attribute Handling for numerically based seaches.
 * Useability: The next problem is, that when we would have lots of different Namespaces the users might not find the right "tree in the wood".
 * MovGP0 01:23, 10 December 2005 (CET)


 * Thanks! I'm still foggy, but that just means I need to learn more about both OWL and Wikipedia. --Russell Thomas 02:56, 10 December 2005 (CET)


 * By the way, my interest in semantic capabilities is a bit different that the mainstream Semantic Web folks (where semantic search is Job #1). I'm mostly interested in various types of automatic processing to create and manage meta-information.  Therefore, I could live with limited OWL functionality as long as the relation types were there.  Or at least that's my current notion. --Russell Thomas 03:23, 10 December 2005 (CET)

Subclassing or Instancing
Question: What are the semantics of this relation? Is it: -- Abraham Fowler
 * "A is a B" == A is an instance of B
 * "A is a B" == A is a subclass of B
 * or both (either/or depending on the case)?


 * I guess the difference is "a" (subclassing) vs. "an" (instancing).
 * (Relation:is a, owl:inverseOf, Relation:is an)
 * Because the one is the exact opposite of the other, we don't need "is an" for expressing the right meaning. In practice many users will want "is an" anyway.
 * -- MovGP0 23:14, 19 February 2006 (CET)


 * Surely "is a" is not the exact opposite of "is an" ? -- Skierpage 03:18, 16 March 2006 (CET)


 * To answer Abraham Fowler's question, if you search for Is a relations, they're all "A is an instance of B". If/when Relation:Is_a gets replaced with Category: support, I think the question goes away, because putting   03:18, 16 March 2006 (CET)

''How about effects of time in this context? Do we say Apollo Rocket is a::Space Vehicle or will it be necessary to have was a or will be relations? How about for things that no longer exist at all, such as Colossus of Rhodes''
 * Exactly. Use "Relation:was a" or "Relation:will be" if you want and sublass them from Relation:is a. -- MovGP0 18:09, 13 May 2006 (CEST)


 * How does the deprecation of Relation:is a affects Relation:is subrelation of?--Joris Gillis 22:58, 14 May 2006 (CEST)

''Maybe there are technical reasons why it must be done early but I think it might be premature to deprecate relations so soon. I think it would be better to let users use simple relations close to natural language, then let the computer transform the data into specific formats and tags. Even from a technical standpoint "Is-a" has a long history going back to Semantic Nets. As it says in Help:Semantics : "Categories can have many different interpretations", and the notion of Category is itself complex. I think yes, still put Category:Page Name at the end of each article but it would be nice to still be allowed to express the data close to the way 2 people would talk about it in casual conversation, with all it's nuances.''
 * Seconded--Joris Gillis 18:32, 15 May 2006 (CEST)

Deprecating "is a": Need to move all class articles to Category?
I understand the use of deprecating Relation:Is_a in favor of using a category. But this brings up the general problem that Wiki users often use an article (which is an instance of a class) to describe something that is actually a class (which should be a category, not an article).

Consider Markus Krötzsch. If he is a Person, then we would use Category:Person. If so then Person should not be an article but should be in the Category namespace. (Wikipedia would have to move all class articles, like Wikipedia Person to the Category namespace. Also note the poorly defined categories for the Wikipedia Person article: "Humans, Personal life, People". The authors are struggling to put something that is already a category into another category.) Alternatively, we can put in a "stub" Category:Person and redirect to the Person article (as I did on this Wiki.)

--Jefft0 21:34, 22 May 2006 (CEST)


 * The problem is that categories are also used to indicate that a resource addresses a specific topic. For example, the wikipage for a scientific article would use the "Semantic Web" category so that it can be found on a query about all the articles about the semantic web, even though the article is obviously not an instance of Semantic web.


 * Typically, making a category a subcategory of another is used both to say that they represent subclasses (e.g. the category Professor is a subcategory of Faculty) or that they represent "parts" (e.g. the category Data mining as a subcategory of Information science in Wikipedia)


 * It seems to me that the problem comes from using the same construct for doing two different things. --OlivierDameron 10:30, 9 June 2006 (CEST)


 * "is a" is no less vague than MediaWiki's Category. If you care about semantics, one approach is where SMW's explicit interpretation of MediaWiki's Category: property in RDF Export is incorrect, don't use Category.  Instead create a relation such as Covers topic::, and then use Help:Import vocabulary to associate this with a well-defined ontology in RDF Export, such as SKOS. -- Skierpage 23:33, 10 August 2007 (CEST)

Is using categories that much better?
Doesn't this paper prove that categories are a rather poor tool to express semantics? Relation:is a might have an ambigous meaning, but aren't categories worse? --Joris Gillis 19:17, 7 June 2006 (CEST)

''Also, apart from SubClass/Member of and Instance of distinctions, how do we express Type Tokens ? See Wikipedia'sClass article for what I mean. Compare with Class here. USoy


 * I'm not happy with this decision to. Cathegories are not the same than owl:Class. Examine a Cathegory like SI units. Well, in that case of handling Categories like Classes, most of the Articles would be correct instantiated as SI Units (with the exception of the plural) and also correct into the Subclasses like "SI base units" and "SI derived units". But the articles "International System of Units" and "Metric typographic units" are be totally right in case of Categories, but would be totally false in case of subclassing.
 * If you ask me, than a Category is just a way of bringing things together wich are related very close to another, so a Human can get a good overview.
 * I've always thought about Relation:is a having as easy to memorize Alternative wich is kind of both, rdfs:subClassOf and rdfs:subPropertyOf. But to categorize something has just the meaning of skos:narrower.
 * Anyway, instead of this we can also use [rdfs:subClassOf]] and rdfs:subPropertyOf. But this more specific concepts will be harder to got understand by not-RDF/OWL-capable Wikipedians. So instead of saying
 * and
 * Wikipedians would be forced to say
 * and
 * Also I've tried in the far past as a Wikipedia-Noob to create dozends of Categories in the german Wikipedia to be capable of better describe all of the Units. Therefore I've created a Categories for Units by Physical Size, Units by Unitsystem in which they are defined, and so on. The effect was an really big edit-war fought within hundreds of Articles!!! I think we should learn on mistakes - even when others did them. So belive me when I say that Categories are a really bad Alternative to explicit Subclassing.
 * &mdash; MovGP0 22:52, 6 August 2006 (CEST)
 * Also I've tried in the far past as a Wikipedia-Noob to create dozends of Categories in the german Wikipedia to be capable of better describe all of the Units. Therefore I've created a Categories for Units by Physical Size, Units by Unitsystem in which they are defined, and so on. The effect was an really big edit-war fought within hundreds of Articles!!! I think we should learn on mistakes - even when others did them. So belive me when I say that Categories are a really bad Alternative to explicit Subclassing.
 * &mdash; MovGP0 22:52, 6 August 2006 (CEST)

In wikipedia, categories are used for tons of different things like "related to an instance of, or subclass of, or related in some other unspecified way" and makes things completely ambiguous. The way I see it, there is a few options.

Use Relation:Is an instance of and Relation:Is a subclass of instead of Relation:is a and have articles for all classes.

Use categories for all classes and mark them as machine-readable with something like is a::class in the category description, and maybe use is an instance of::Category:categoryname and is a subclass of::Category:categoryname

Use categories for things that do not have pages and are concepts that need not necessarily have an article. For example Electrically active cell is a class from Open Biomedical Ontology's Cell Ontology, but I'm not sure if it's the best title for an encyclopedic article by itself and could instead be Category:Electrically active cell.

Fuelbottle 14:14, 7 September 2006 (CEST)