[Top] [All Lists]

Re: [ontolog-forum] using SKOS for controlled values for controlled voca

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>, "John F. Sowa" <sowa@xxxxxxxxxxx>
Cc: Yorick Wilks <Yorick@xxxxxxxxxxxxxx>, Graeme Hirst <gh@xxxxxxxxxxxxxx>
From: Simon Spero <ses@xxxxxxx>
Date: Mon, 11 Oct 2010 13:36:21 -0400
Message-id: <AANLkTiko2cyyzuT6_vptsDj+j9rbybsoB=9-ha6Sye6v@xxxxxxxxxxxxxx>
On Sun, Oct 10, 2010 at 11:38 PM, John F. Sowa <sowa@xxxxxxxxxxx> wrote:
Amanda and Leo,

The distinction between natural language terminologies and formal ontologies is clear in the abstract.  But when we start using ontologies to analyze NL texts, we run into very serious issues that blur those distinctions.

John is right on the ball here, but I'm not sure he's playing the right game (had to get a Wittgenstein allusion in there somehow).  

Controlled Vocabularies are Controlled, in the same way that Controlled Natural Languages are Controlled.  Just as documents written in ACE (Attempto Controlled English) can be read by English speakers with minimal training, Terms in a Controlled Vocabulary can (or at least should be) readable by native speakers. 

However, the constraints that are placed on the Terms permitted/the subset of syntax recognized require a certain degree of training before new terms can be added, or new sentences written.

All controlled vocabularies must control for synonymy; all labels for a specific concept must be attached to that concept, and the same label cannot be used for multiple concepts 

Controlled Vocabularies that allow Hierarchical relations must follow other constraints: the most important one is that the relationship must always be true. That means that if Term A has Broader Term B, everything that is about A must necessarily in some fashion also be about B. [Term == Concept in KOS speak].   

The KOS standards only allow for a very limited set of hierarchical relationships, because they require the relationship to always be true.  Where the standards are followed correctly, it allows subject trees to be rolled-up , avoiding the need for the loss of specificity caused by things like the upward posting mentioned earlier.  Of course, when they're not followed, you end up with messes like the Library of Congress Subject Headings (everything about Doorbells is necessarily also about Eschatology? This is the way the world ends - not with a bang, but with a ding-dong?)

I can't remember the precise quote, but one of the CRG members  (don't have reference to hand) defined the broader relationship as not being to allowed to complain if you asked for documents about B and also got documents about A. 

The important thing to remember is the KOS Concepts are essentially intentional-with-a-t. The entities and relationships are  intensionally-with-an-s those of aboutness; the extensional semantics are defined in terms of ``Documents'' to which the Concept could be applied.   They are not strictly lexical, and are people who have looked at SKOS for use in NLP and CL  have opted not to use it because of a lack of fit with their needs from a Lexicon .

There are some ontological inferences that can be drawn, but the precise inferences are specific to how the  KOS was constructed; 

Simon //  "Word and Subject"? It's ok Van - I'm only pretending there are intentions.


Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (01)

<Prev in Thread] Current Thread [Next in Thread>