ontolog-forum
[Top] [All Lists]

Re: [ontolog-forum] XML tags as natural language words

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: "Neil Custer" <neil.custer@xxxxxxxxx>
Date: Mon, 15 Dec 2008 14:04:47 -0600
Message-id: <7a5bda850812151204o7adc4d53kf9e7f6d636984b9b@xxxxxxxxxxxxxx>
I've noted that some efforts to use structured schema naming and design rules such as those used in the UN/CEFACTS' CCTS and subsequently OASISs' UBL do well at retaining precise meaning in specific contexts, in the case of UBL for those business transactions whose data must be exchanged between business partners.  In lieu of a specific ontology that meets with across-the-board agreement for terminology used, the data representing a specific business need during a specific business transaction suits their needs just fine. 

I am not optimistic that a "general ontology" using formal or NL definitions can ever be reached using 100% agreement as the determiner of success. 

This is just an idea, but I think that perhaps more specific contextual definitions where specifically-defined domains and ranges of values in a particular context may be obtainable.  If specific ontologies could be agreed upon 100% for those of "general nature of content", "domain + range", and "context", then relate those together as Tania Touradache has in her engineering thesis, then perhaps you may have a viable direction.  But I would agree with anyone that believes 100% NL agreement can ever be achieved, especially when used only as an annotation.

If it would be helpful, in light of John's comments, I might also suggest that if one could develop a framework whereby he could represent specific meaning of a term without using any other words at all, this would be the thought processes one must use in order to determine a solution to these problems.


On Sat, Dec 13, 2008 at 4:58 PM, John F. Sowa <sowa@xxxxxxxxxxx> wrote:
Ravi,

Those are in the same category as other natural language resources:

RS> Where would we put Vocabulary and Thesaurus efforts, would they
 > help in better tagging XML or better NL definitions helping efforts
 > in ontologies, or both?

There are several problems with tags that come from any resource
defined in any NL (e.g., terminologies, vocabularies, etc.):

 1. The tags (or words) are defined by NL statements, not by axioms
    in some version of logic.

 2. Almost all the people who annotate texts using those resources
    think in terms of some NL definition.  (This point is true
    even if the terms are defined in logic, but the annotators
    don't understand all the implications of the formal logic.)

 3. Annotators who are not trained lexicographers tend to get
    error rates of about 22% to 33%.

 4. Even professional lexicographers have inter-coder disagreements
    of about 4.5%.

And as Yorick W. and others have pointed out, even if the tags
are defined by formal axioms in some sort of logic, the definitions
tend to vary over time as the axioms change.  Yorick mentioned Cyc
as an example, for which the definitions of the formally defined
categories have changed over the past 24 years.  Therefore, those
categories are no more precise than typical NL word senses.

John



--
Neil

Life is Great!-- It sure beats the alternative!!  
So live better - live longer!  See our exclusive health and nutrition products at:  
<a href="" href="http://ncuster.qhealthzone.com">http://ncuster.qhealthzone.com" /a>



_________________________________________________________________
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>