Re: THEORY: What IS language anyway?

Gary Shannon wrote:

> We each construct a mental model of the world. When we
> "remember" an event we do so by simulating a recurance
> of that event within our mental world model.
Great spirits always meet.  ;-)  Shaquelingua is based upon a theoretic
world model not very different from the way you re-analyze your example
sentences, I mean movements of "objects" from one metaphorical
space/entity to another. However, in Shaquelingua, these "objects" can
be either objects (this or that) or qualities (like this or like that).
I called this the Flux.

A class of particles describe the nature of the Flux, i.e. if the moving
"object" is an actual object or a quality, and what arguments are needed
to make a valid sentence.

> Even conveying static information, such as "Einstein
> was a physicist." can be interpreted as a command to
> the world simulation, in this case making an attribute
> assignment:
In Shaquelingua, this is an equative flux.

	Flux(
		Type of Flux = equality with quality,
		Quality = physicist,
		Initiator = Einstein,
	);
	Einstein equals quality "physicist".

In a way, you can say that my computer language uses typed variables.  ;-)

> For example: "John gave a book to Marsha." compiles
> into the command:
	Flux(
		Type of Flux = dotation of object,
		Object = book,
		Receiver = Marsha,
		Initiator = John,
	);
	John put the book into Marsha's inventory.

> "Marsha took the book from John."
	Flux(
		Type of Flux = acquisition of object,
		Object = book,
		Transmitter = John,
		Initiator = Marsha,
	);
	Marsha removed the book from John's inventory.

> "Marsha stole the book from John."
	Flux(
		Type of Flux = attribution of quality,
		Quality = stolen,
		Referential = John,
		Receiver = book,
		Initiator = Marsha
	);
	Relative to John, Marsha add "stolen" to the qualities of the 			book.

I made a different analysis but there's no big difference in between
taking and stealing. The book is transfered from John's inventory into
Marsha's. I could build the same sentence and add the circumstantial "by
robbery" to mean that it was a theft.

> The key, then, to having a computer "understand"
> language is programming the computer to compile
> sentences into simulation commands and then to carry
> out the simulation in the computer's own world model.
That's a good start. But now analyze the following sentence:

John pulled Marsha's leg.

Let's imagine different situations:

John and Marsha are WWF fighters in a ring.
John and Marsha are floating in a zero-G environment.
John is Marsha's baby boy trying to catch her attention.
John and Marsha are discussing in a bar.
Marsha is the doll of John's sister.

To analyze a sentence, the context is important. There is always a
context, not only a factual context (who, what, where) but also the
physical surrounding of the sentence (where it is read or heard). Even
when there's no apparent context, our experience provides one. We know
the laws of physics (how objects behave or react), we know how people
usually interact --especially if we personally know them. Our memory is
   also a huge database of facts. We are even able to understand
sentences with missing words.

The cat chases the...

How many "chasees" did you find at the end of the sentence?

That's what a computer needs to "understand": a memory based on
experience or the closest equivalent we can provide.

--
==================
Remi Villatel
maxilys_@_tele2.fr
==================

Reply