[MUD-Dev] Re[2]: [MUD-Dev] Languages (slightly offtopic, was Text Parsing)

[Travis S. Casey]

On Wed, 2 Jun 1999, Matthew Mihaly wrote:
> At 08:42 PM 6/2/99 -0400, Travis wrote:

> >That is the difference between *prescriptive* and *descriptive*
> >grammar.  A prescriptive grammar, which is the kind of grammar you're
> >exposed to in grade school, is some group's idea of the way the
> >language *should* be spoken.  A *descriptive* grammar, which is what
> >linguists work with, is a description of the way the language *is*
> >spoken.
> 
> Descriptive grammar seems, to my uneducated self, to be impossible to
> define precisely and thus absolutely useless in terms of a mud parser. If
> grammar is the way language is spoken, then any set of words that is meant
> to express an idea can be considered English. If descriptive grammar is
> what is spoken, then by speaking it, it becomes proper grammar. I realize
> this is a bit silly, but how many people does it take to become correct
> grammar?

That's an open question.  However, it's the same sort of question that's
faced in lots of other situations in the real world.  For example, what's
a species?  Are wolves and dogs the same species, since they can
interbreed?  How do you define species for animals that you can't observe
the breeding of -- e.g., dinosaurs?

Let's say you were researching TV shows.  How do you go about deciding
what defines a comedy versus a drama?  One way would be to find "experts"
and ask them.  Another would be to take a randomly selected group of
people, show them the shows, and ask them whether each show is a comedy or
drama.  Take the shows that they show the most agreement on, and then
start analyzing those to determine what's different about them and what
they have in common.

The same thing applies to languages.  Let's say that you find a previously
undiscovered tribe in the Amazon rain forest, and you want to produce a
grammar and vocabulary for their language, so that other people can learn
it before coming to deal with them.  

There aren't any authorities you can consult about their language -- they
don't have formal schooling, and they don't have a written language, so no
one's ever tried to codify their language before.  What do you do?  (We're
assuming that you can speak with them -- either you've stayed there a
while and started to learn their language, or you've found an 
interpreter.)

The logical thing to do is the same sort of thing that we described above
for TV shows:  take samples of their language, e.g., by recording them in
casual conversation, and then take a group of them, play back those
samples for them, and ask if they're properly-formed.  If they're not, ask
what's wrong with them.

(People can tell if a sentence is "properly formed" without having a
formal grammar.  For example, any native English speaker will tell you
that something's wrong with the "sentence" "Go Joe market did to."  Or,
they can tell you that "I mean -- that is -- I'm trying to say -- oh, the
heck with it" isn't a proper sentence.)

It'll take time, but you will be able to figure out their language.  What
you get out of it will be a descriptive grammar -- you're describing how
their language *is*, rather than how someone says it *should be*.

Now, do the same thing with speakers of English.  Since English is spoken
over a far wider area, you'll find a lot more variation -- but you will
find some things that are common in all variations of English, like the
basic SVO order for statements.  

> I don't have a great understanding of programming, but it seems to me that
> the only way to code a parser is to make it use an objectively defined set
> of rules. Whether those rules come from Oxford or your imagination, you are
> still defining your in-game grammar just as rigidly. I don't mean this in a
> deragatory way. I just don't see any other way to do it.

I don't either.  However, you can objectively define a set of rules in a
pragmatic way -- that is, find a set of rules that fits what's actually
being said.  That set of rules is a descriptive grammar.

> It also occurs to me that the examples you provided were all
> speech-oriented. Speech and text are two different things, and if I picked
> up a history book that was written the way people speak, I'd throw it into
> the garbage. I might not laugh at someone who said something to me like
> "That dog ain't got no fleas", but I certainly wouldn't take that person
> seriously. Likewise, I try to make the output in my game use proper grammar
> (to me, that means Oxford grammar). As far as input goes, I'm happy forcing
> them into a rigid structure. I guess while a parser that can accomodate
> constantly changing dialects would be interesting, I don't see the point in
> terms of MUD development. 

*Sigh*.  I never said anything about it being useful for mud
development... I was just pointing out that Caliban's statement that
people can't speak their own languages "correctly" was BS.

The English grammars used in schools are meant to teach people to use a
certain form of English that people have decided is "better" than other
forms.  It has as much to do with what English really is as the standards
for judging dogs at dog shows have to do with what a dog is.

At any rate, this is getting far off-topic.  Suffice it to say that I
agree with you that a mud parser needs to be based on an objective set of
rules.  I also agree that, if you're trying to use English-like input,
it's best to use Standard English as your model -- even if it's not
everyone's native dialect, anyone who's logging into a mud has probably
been exposed to it.

--
       |\      _,,,---,,_        Travis S. Casey  <efindel at io.com>
 ZZzz  /,`.-'`'    -.  ;-;;,_   No one agrees with me.  Not even me.
      |,4-  ) )-,_..;\ (  `'-'
     '---''(_/--'  `-'\_) 



_______________________________________________
MUD-Dev maillist  -  MUD-Dev at kanga.nu
http://www.kanga.nu/lists/listinfo/mud-dev