Re: [ontolog-forum] syntax & semantics

[Jawit Kien <jawit.kien@xxxxxxxxx>]

On Mon, Apr 27, 2009 at 11:15 AM, Richard H. McCullough <rhm@xxxxxxxxxxxxx> wrote:

According to the dictionary,

 

    the syntax of a language is

    the grammatical rules for combining terms

    into phrases and sentences of a language.

 

    the semantics of a language is

    the relation of terms to their meanings --

    the things which those terms denote/refer to.

 

In a most unfortunate perversion of definitions,

probably due to the influence of Noam Chomsky,

"formal semantics"/"model theory" is concerned

with the syntax of a language, independent of

the semantics of a language.  Formal semantics

looks at the properties of the syntax of a language

which do not depend on the actual semantics of the

language.

Syntax is essentially, the process of dealing with forms
of things and rigourously describing those forms so they can be
recognized by a parser, or generated by some process.
As I see it, the phrase "formal syntax" is just a bit redundant,
but maybe it just means that the rigour used to describe the
forms that expressions/sentences of the language might match
is important.  So while the syntax of the language is focused
on the way someone might type in or "tell" the computer some
information, if it is an informal syntax, it might be described with
generalized terms. 

I can imagine an informal syntax for a language as being
"the words people use to describe the sounds animals make"
with sentences in that language (for English speakers) being "moo"
"oink", "meow", "ruff", "bow-wow", "neigh", "baa", "whinny", etc. 
For non-English speakers, there would be other words, which might
be interesting to know, and those words would be in this language too.

But this isn't a FORMAL syntax. Frankly, I don't know if this language
has much of a formal syntax, it might just be a list of words. (a vocabulary)
I guess if you say a dog "arfs", you could say it "arfffs" so you could
make a grammar that has the possibility of multiple "f" characters,
so it could be possible to describe this language, which I initially
would say is just requires a vocabulary, to usefully having a grammar,
and thus being able to be described formally.

Formal semantics is formal first of all, then semantics after that.
so while a semantics is some kind of meaning, if it is to be a
formal semantics, then it must have a "form" that is associated
with it. The form CAN'T be about the way that the sentences
of the language are defined, because that would be the syntax
of the language.  The form must be something different.

One possible thing that the formal semantics could be about,
is that the forms that it deals with would be the forms that
logical statements could take.
To explain a little more, I can imagine a logical theory that
had a few predicates like "makesSound" or "soundsLike", and
then a set of identifiers for animals, and a set of sentences from
the syntax of the Animal Sounds language mentioned above,
the logical theory would contain logical structures to describe
this particular aspect of the language.  This would be formal
because it would deal with the FORMs that we could use
(i.e. the shapes that the extension of the predicates might make),
and it would be semantics, to the degree that the logic expresses
the "meaning" of which the various syntactic forms might have.

So to be specific, I imagine there is a predicate "makesSound"
would establish a link between an element of the extension of
whatever "Animal" describes, and the sentence in my Animal
Sounds Langauge. I believe this description constitutes the
intension of the predicate. Some of the extension of the
set would be:

makesSound(Dog,"arf")
makesSound(Dog,"bow-wow")
makesSound(Dog,"ruff")
makesSound(Cow,"moo")
makesSound(Pig,"oink")
makesSound(Hog,"oink")
makesSound(Mouse,"eek")
makesSound(Mouse,"eeek")
makesSound(Mouse,"squeek")

I think this would be expressed in Conceptual Graphs in several ways
I'll try to use different variants for each of these below: (John Sowa, am I right?)

[Dog:] -> (makesSound) -> [:"arf"]

[Dog] -> (makesSound) -> [:"bow-wow"]

(makesSound [Dog:] [:"ruff"])

[Cow] 1-> (makesSound) 2-> ["moo"]

[Pig] -> [Hog] -> (makesSound) -> ["oink"]


the other predicate might be "soundsLike", which links
two Animals or two sounds together.

soundsLike(Pig,Hog)
soundsLike("arf","ruff")
soundsLike("arf","bow-wow")

I could even add a rule, making this a logical theory that
could be used for inferencing:

ForAll S,Q
 if (soundsLike(S,Q) AND (soundsLike(S,R))
then soundsLike(Q,R)

which could be used to infer
soundsLike("ruff","bow-wow")

This is kind of like a transitive rule, but I'm not sure if
it technically is, could someone tell me?

By the way, interesting links to follow:

Wordnet (using SUMO's page) has
http://sigma.ontologyportal.org:4010/sigma/WordNet.jsp?word=animal&POS=1

SUMO has
http://sigma.ontologyportal.org:4010/sigma/Browse.jsp?kb=SUMO&term=Animal

 

That is why I associate

syntax with "possible meaning"

and semantics with "real meaning".

 

I'm sorry, this seems arbitrary and something I would have
to file as "specialized words that only Dick McCullough has
a clue what he means".  The word syntax doesn't mean possible
in any way to me, and syntax is not tied to "meaning" in my mind.
If you think "real meaning" is the meaning inside someone's head,
fine, but I don't have any clue what the forms that the meaning inside
my head has, but if I express those forms using logic, then I have
a hope that a computer can come up with other forms using logical
inference that correspond to some other meaning inside my head which
I would think is a consequent of the meanings that I originally had.

If you think having the computer come up with logical inferences
is not useful, then it isn't necessary for you to describe what
expressions in your language map to in a formal logic, such as IKL,
or Common Logic.

I personally think it is useful for the computer to come up with
these logical forms ("formal semantics") for the syntax I express.
I would like to see if it makes life easier for me, and helps me
organize my thoughts better.

 

You (and Pat Hayes and Chris Menzel) are very

focused on syntax.  Whenever I mention some

issue relating to the semantics of mKR, you change

the subject to [the "missing" documentation of]

the syntax of mKR.

I think you are confusing semantics with syntax.
I agree with Pat, Chris, and John that just having words
on a page is not as useful as having words on a page
that a computer can use to draw conclusions and make
inferences.  That is the point of an ontology, and that
is the point of this mailing list/forum.
 

 

Let's not lose sight of the fact that semantics

logically precedes syntax.  As a language evolves,

there is an interaction between syntax and semantics,

but syntax is the servant of semantics.

How do I know what you mean by the term
"logically precede" ?
proceeding normally is tied to motion along a path from
a start position to an ending position.
prEceeding normally means that some thing is moving
ahead of some other thing whilst both are proceeding.

Thus to try to understand what you just said,
if something "logically precedes" something else,
we are basically making an analogy or metaphor.
Since "precedes" is a verb,  we need a verb which
has some kind of directionality associated with it,
and which has a metaphorical start position, path,
and ending position. Moreover, this verb must be one
where "logically" makes sense when used as an
adverb to modify that motion-verb.

One way of viewing "logic" is as a way of describing
an inference process.  Metaphorically, we can then
identify the motion-verb is "inferring", or "deducing"
the path as the steps taken in a logical proof,
the start position as the facts that were explicitly
stated to be true at the beginning of the proof-process,
and the end-position as the facts that were explicitly
stated to be true at the end of the proof-process.

This metaphor seems to be the one you were thinking
of, but now I need to have two things that I identify
with "motion", and where one can be said to be
"preceding" the other.  The beginning facts of the
proof-process certainly exist before the ending facts,
so they could be said to "preceed" the ending facts.
If I identify the beginning facts with "semantics"
and the ending facts with "syntax", I end up in a
conceptual muddle. 

Since I know you have said something in the past
about semantics being what you "mean" by a mKR
statement, are you just saying that what you think
in your head must be filtered through some logical
process (presumably in your head) which will
eventually yield the actual symbols you write down
on on paper i.e. the syntax?

If this is what you mean that "semantics precedes
syntax", how does the computer even come into
play until the last, where you use the computer to
record what you were thinking, instead of writing it
down on a piece of paper?

Inquiring minds want to know,

JK

 

 

Dick McCullough
http://mkrmke.org

_________________________________________________________________
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
 

_________________________________________________________________
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)