ontolog-forum
[Top] [All Lists]

Re: [ontolog-forum] Semantics of Natural Languages

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Phil Murray <pcmurray2000@xxxxxxxxx>
Date: Mon, 18 May 2009 18:27:14 -0400
Message-id: <4A11E0C2.7020707@xxxxxxxxx>
John --    (01)

Thanks for the pointers to resources about the "semantics of natural 
languages." (And thanks to Azamat, too, for his response.)    (02)

Schlenker's paper ("Semantics") was challenging but impressive and 
thought-provoking for someone without an education in the field.    (03)

But the last sentence in this extract from his initial summary seems, 
er, wrong.    (04)

    ... if we provide any speaker with a syntactically well-formed
    sentence S, together with a sufficiently detailed situation (i.e. a
    description of the way the world is), the speaker should in
    principle be in a position to determine whether S is true or false.
    It is clear that speakers have this ability for infinitely many
    distinct sentences and situations, which they could not all have
    memorized. *Therefore they must have access to certain rules that
    allow them to compute the truth conditions of complex sentences on
    the basis of memorized facts about their smallest component parts.*
    [emphasis added]    (05)


This is not a given -- at least, not when stated precisely in this way. 
That is, it's not true that speakers (and listeners) break sentences 
down into their smallest component parts in order to create (or 
interpret) their meaning. In most cases, anyway. This is obviously not 
true if we claim that one concept = one word. (That's a trivial 
criticism, and I think that was not Schlenker's intent.) However, it 
does appear that Schlenker is claiming that a speaker (or listener) may 
interpret any number of multi-word phrases as elements of meaning that 
are reusable in any context. And that seems clearly not to be the case 
in reality. In some cases, the meaning of a phrase is at odds with other 
instances of the same phrase in different contexts/different sentences 
... and at odds, as well, with the known meanings of its component 
words, even in their most metaphorical senses.    (06)

Sometimes we can and do express meaning directly and unambiguously with 
natural language. "Sit!" is certainly an unambiguous command in English. 
"This person is genetically identical to that person," is probably an 
unambiguous statement of reality for most English speakers.    (07)

However, metaphor (one form of analogy?), intent (pragmatics), and the 
cumulative associations with all our previous experiences of using 
language to represent meaning (including the successes and failures of 
our attempts to communicate in natural language) together create a 
verbal miasma that humans are uniquely equipped to decipher ... although 
often quite poorly, one must admit. It seems to me that current models 
of KR are probably ill-equipped to solve this problem any time soon, in 
spite of the thoughtful work by ontologists (including many in this 
forum) and linguists.    (08)

In fact -- from the perspective of putting meaning to work in real life 
(and in real work) -- we seem to get into trouble by focusing so 
intently on this effort to deconstruct natural language into meaning. 
(The important exception is the application of ontologies to 
interoperability of software.) It would be more productive in the near 
term to look for specific types of information in common communications 
that would benefit directly from being replaced by richer, useful, 
formalized representations of meaning that are largely *isolated* from 
natural language, but which serve as a complement to natural language.    (09)

We don't build arithmetic on verbal approximations (and the contexts in 
which they occur). We don't try to make something usable and functional 
out of "a bunch of," "several," "a barrel of," or "a pinch of." (At 
least, we no longer do that.) Because that may be sufficient for an 
informal transaction or baking a cake on Tuesday, but it's not at all 
sufficient for commerce or the production of goods. Why should we 
continue to use [natural language-based] approximations of meaning for 
the most critical aspects of work?    (010)

We *do* fundamentally understand that we have to make meaning explicit. 
We see it every day in relational databases, product catalogs, graphic 
assembly instructions and blueprints, process designs, and other areas 
of human activity. And there are useful aspects of meaning not well or 
efficiently represented by natural language. We certainly haven't spent 
enough time thinking about those things and putting them to work, IMHO.    (011)

The transition to a "semantic" or "meaning-based" approach to knowledge 
work is not simply likely; it is inevitable. But we'll need practices 
and tools that are both (1) more systematic across enterprise activities 
and technologies and (2) more immediately practical for a wide range of 
users within those activities ... but which do not become an end in 
themselves.    (012)

Phil Murray
Chief Knowledge Architect
The Semantic Advantage
blog: http://semanticadvantage.wordpress.com/
web site: http://www.semanticadvantage.com
401-247-7899    (013)



John F. Sowa wrote:
> Many of the discussions in this forum have addressed issues related to
> the differences between formal ontologies and the semantics of natural
> languages.  For anyone who would like to see what linguists of various
> persuasions are saying about semantics, following is a list of papers
> and PhD dissertations on semantics that are available for free download:
>
>     http://semanticsarchive.net/sem-bin/browse.pl
>
> This list is in reverse chronological order from 2009 to 2000.
>
> Most of the papers address very narrow and detailed issues, but some of
> them are more general.  For a survey of formal semantics written as
> an encyclopedia article, see
>
>     http://schlenke.free.fr/Semantics.pdf
>
> I have very serious doubts about any treatment of NL semantics that
> does not come to grips with the issues raised by Wittgenstein's later
> philosophy, but this article is a reasonable summary of what formal
> linguists are doing.
>
> Following is an article that addresses issues of context:
>
>     http://semanticsarchive.net/Archive/jVhNzFiM/QPRS.pdf
>
> The following article raises a fundamental issue for ontology:
> do 'Cheeto' and other trademarked terms refer to a natural kind
> or an unnatural kind?
>
>     http://semanticsarchive.net/Archive/DU2ZmMwN/copyright.pdf
>
> This list of articles illustrates the wide range of issues and
> the unlikelihood that anyone is going to develop a unified
> universal ontology that can address all of them simultaneously.
>
> John Sowa
>
>  
> _________________________________________________________________
> 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
>  
>
>       (014)

_________________________________________________________________
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    (015)

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