Re: Throwing out the tree-structured grammar (SF Xenolinguistics FAQ)?

On 6/11/05, Ray Brown <ray.brown@...> wrote:
> 
> 
> No one has mentioned the FAQ author's bizarre "hash". While this may,
> properly executed, be a convenient way for a machine to store and retrieve
> data quickly without (in theory) any searching, how would it contribute to
> direct communication?

[snip]
> 
> My impression is that the author of the FAQ has not really thought through
> what s/he has written. IMO without the attempt to describe 'such horrors'
> it does not really say anything more than "Alien grammars may be quite
> different from any grammars of our familiar earth languages."
> 
Well, in general I found the document to be pretty entertaining, but I too 
would doubt that the author really has any further thoughts on the matter of 
non-tree languages. The reference to a hash grammar is most likely a 
strained analogy to trees -- hey, both trees and hashtables store grammar, 
so why not a hash grammar instead of a tree one? The analogy only weakly 
holds, since it's not primarily the data *storage* properties of trees that 
human language takes advantage of. I suppose we could argue forever about 
what property it is (other than recursiveness) that we take advantage of, 
but let's call it data *organization*. Hashtables store data efficiently, 
but they are not really meant to organize it. Underneath it's just (for 
example) a list of lists, and a quick way to get at any individual item 
provided one knows the property it's hashed upon.

But let's play...

We have an alien species called the Plogl (pKOgK\), who are slow and 
deliberate thinkers but can communicate huge amounts of information at a 
singe blow by expelling information-rich spores. They can communicate huge 
one-dimensional arrays of chemically-encoded "symbols" very quickly, but are 
slow at parsing them. (Incidentally, they are quite slow at constructing 
them, as well.) Each symbol encodes at least the following information: a 
thing, property, or event, its proximity to the broadcaster, its gender 
(sweet, salty, bitter, or sour) if applicable, and its semantic and 
pragmatic roles in the "sentence". Huge amounts of data are included in each 
"broadcast", and much of it is not immediately necessary, but a sort of... 
well, things to explore, savor, and ponder over in the several days before 
another "broadcast" arrives.

Symbols are not presented in "order", since they are not consumed in any set 
order and, anyway, the information considered interesting may differ between 
two Plogl...es? en? i? Rather, they are hashed on the value of their role in 
the sentence. (A sentence is an array of usually less than one thousand 
symbols. A broadcast consists of usually less than 300 sentences, themselves 
hashed along similar lines.) 

So upon isolating an interesting sentence, the receiving Plogl might work 
out the position of "agent" (In her language, "agent" hashes as 1,039 
modulus the length of the sentence, skipping 29 spaces each time in case of 
a collision). After also figuring out the patient, the beneficiary, the 
maleficiary, the event, and what the situation smelled like, she might 
decide to further investigate the patient. So she hashes on "patient's 
possessor", and "patient's possessor's possessor", "patient's source", "what 
the patient smelled like", and "what the patient's possessor smelled 
like"... etc.

(You'll notice of course the recursive nature of these roles, which suggests 
that some treeness or graphness is going to be necessary somewhere in the 
system. Either that or there's a stock of some n roles beyond which no new 
roles can be expressed. This goes back to the fact that hashes themselves 
don't organize, in the above sense of organize. If we want organization, 
something else has gotta provide it.)

Anyway, there's a hashy language for ya. The hashtable structure solves the 
fundamental problem of Plogl communication, which is that the Plogl needs to 
sift through a huge 1-dimensional array of information, some of which this 
particular Plogl will find interesting and some of which she won't. There 
are other ways of solving this problem, but by whatever invisible hand 
guides the evolution of the Plogloi, hashing is what they've developed.

-- 
Patrick Littell
PHIL205: MWF 2:00-3:00, M 6:00-9:00
Voice Mail: ext 744
Spring 05 Office Hours: M 3:00-6:00