[Top] [All Lists]

Re: [ontolog-forum] Standardization of ontologies

To: <ontolog-forum@xxxxxxxxxxxxxxxx>
From: "FERENC KOVACS" <f.kovacs@xxxxxxxxxxxxxx>
Date: Mon, 6 Jul 2009 12:25:43 +0100
Message-id: <8BC8727F6A63461F954B67A2856442D6@Swindon>

Hi All,

In my view Formal Logic and formal (what you call semantic) analysis (syntax parsing) of NLs, as that of the English language as it is done now by ontologists and MT champions is a misleading practice, rather than reliable scientific (sensible) endevaour with respect to creating any imperoperability between ontology terms. The reasons are simple and easy to identify as i am trying to illustrate below.


In using natural languages most communication actors heavily realy on your knowledge of the environment, that is the meaning and the context of the verbal input available only through staged in a number of layers, including the text itself, various definitions, the knowledge of the  participants, the time of reading, etc.

They also rely on our basic attitude to seek sense and correct whatever deficiencies need to becorrected to change a verbal input to make sense. This stems from the confidence of normaly bred children in the adult world and that life is something with a purpose. So any disambiguities like this:   ?see results of a search in calendar view? would be rephrased as 1) please check the calendar view to see the results of your search or  2) check out the results of a search performed in calendar view subject to the knowledge of the recepient.


In MT and the software used by translator agencie where TMs are produced by mechanically chunking texts taken from textual data bases for instance, and decontextualizing them just as the headwords are in dictionaries, nowadays you tend to get very funny fragments that make no senses at all.


First, some of the languages are non indo-European, so for those speakers any such decompositional exercise which produces "of the" in a cell of a table is a nightmare, and whoever assembles them again will get nonsense. Similar story concerns parameters. In a single sentence you may have as many as three different parameters all marked as.% (like in  %s on %s is %s) in that order. Now unless you tag the parameters, you will suck there in translation or real semantic interpretation. English speaking journalist take pleasure in creating paradox sounding headers and titles, such as

About 235,000 Americans have their portraits taken by professional
photographers every day.
We guarantee to replace all defected parts.
Japanese Race Appears Reduced to Three.
He likes to watch the sun go down on horseback.
Congress Votes for Running Trains Over Union Workers.
City Gets $250,000 to Combat Drug Program.
Keep Drinking Water From Sewage.
Doctors discovered his ankle had been fractured in five places during
She was one of eight children born to Walter and Irene Smith on
Feb.13, 1892.
The building is a home for underprivileged children under
Food Stamp Recipients Turn to Plastic.
Police Officers Complete Crime Course.
Lung Disorders Meeting.
and the best f all of
US journalism

Florida Reporter Completes Sentence.
And so on.

Such utterances illustrate that semantic analysis does not work through syntax only, and the underlying FOL as used today is incorrect, incomplete and inadequate, just as Formal Logic is for the semantic analysis of natural languages. I do not doubt that it is a bitter tablet, but surely one day you will have to swallow it.No to mention some other basic facts that the human mind is interpreted to consist of reason, emotion and will, of which will is in control and whatever reason or reasoning is to produce may be strongly transformed by the impact of emotion.




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>
  • Re: [ontolog-forum] Standardization of ontologies, FERENC KOVACS <=