Re: Sim Conlang (Re: When is plural applied?)

On Wed, 14 Apr 2004 11:58:11 -0700, Gary Shannon <fiziwig@...> wrote:

>--- David Peterson <ThatBlueCat@...> wrote:
>> Gary wrote:
>>
>> <<What we need is something like "Sim City", a
>> computer
>> program that could be programmed to simulate a bunch
>> of little people running around speaking our conlang
>> so we could watch evolve for a few simulated
>> centuries. ;-)>>
>>
>> You have NO idea how many times I've envisioned
>> exactly this!!!   To me, it'd
>> almost be like there were levels.   You'd enter a
>> certain amount of words and
>> morphological info then hit "start", and the level
>> would start.   There'd be
>> a bunch of people (e.g., someone buying food from a
>> shop, the store keeper, a
>> policeman, a woman trying to park her car,
>> passersby, a guy at a magazine
>> stand, etc.), and they'd just kind of act randomly
>> and try to interact.   The
>> second there was a problem (i.e., they'd try to
>> interact, but couldn't, based on
>> the input), one of two things would happen: (1)
>> they'd inovate something, and a
>> log would record the inovation, or (2) if it was too
>> serious, the level would
>> stop (i.e., you'd lose), and you'd be told exactly
>> what went wrong.   That
>> would be AWESOME!!!   I would pay twice the
>> suggested retail price for it.   I
>> wish I were rich, computer-brilliant, and had
>> nothing else to do: I'd get right
>> on this.`
>>
>
>Considering the difficulty of doing ANY kind of
>natural language processing by computer (from machine
>"understanding" to machine translation) I'm afraid
>this is a project that is beyond the reach of current
>technology.  At best it would have to use some very
>stylized "grammar" and rigidly constrained set of
>things that could be talked about.  That wouldn't
>leave much room for any kind of innovation or
>naturalistic evolution, especially since the
>programmer would have to pretty much pre-plan any
>evolutionary steps that were to take place, and that
>defeats the whole purpose.
I've thought of doing exactly this many times before too, but I've reached a
similar conclusion: simulating realistic semantics would require better
machine understanding of natural language than we have with current
technology.  But there are other aspects of language which should be easier
to simulate.  Sound change is probably doable.  In fact, once upon a time,
there was a random phonology generator on the web -- http://www.bol.ucla.edu/
~smithma/Random/random.html -- but it seems to have vanished; sound change
should be a similar sort of thing.

Alex Fink