Hi,

> My point is not that grammar isn't important. My point is that parsing
> should not be the first step. The first step is to identify the
> semantic units (words and suffixes). The second step is to learn their
> meanings (associations with words and possibly non-verbal grounding).

That is the sorta thing we're aiming to do with our game-world and robotic
embodiment for OpenCog.  Have the system ground words in what it perceives...

That is indeed how humans proceed with language learning.  But it's
not necessarily
the only way for an AGI system to do things.

Humans also start with gesture before speech, and with speech before text,
which is not the approach you're taking.  So you are also willing to
take a route
differing from human development when you find it convenient!!   You
are making your
own intuitive judgments about which parts of human language processing
and development
are important to emulate in your AI system...

> Most of the meaning of a sentence is the sum of the meanings of its
> words.

That is what folks doing statistical language processing always say.
I think it's a boring,
and not very useful statement.  I think you are creating a bogus
theory of semantics in order
to match the limitations of the algorithmic toolset you've found most
useful for text compression.
The simplest, one-year-old-child aspect of the meaning
of a sentence is the sum of the meanings of its words....  Summing
word meanings gets you
a very, very small distance toward interpretation of complex sentences
like the ones we
typically use on this mailing list.

>Word order and grammar plays a lesser role, but we can't ignore
> it entirely. I realize that "Alice hit Bob" is not the same as "Bob
> hit Alice". But I think the most important role of grammar is error
> correction.

This is a profound conceptual error on your part....  There is a huge
body of linguistic theory
contradicting this idea, see e.g. Jackendoff's book

http://www.amazon.com/Foundations-Language-Meaning-Grammar-Evolution/dp/0199264376

for some elementary education in cognitive linguistics ;p ...


> How does the link grammar parse the following?
>
> "I saw the star with the telescope"
> "I saw the star with the hat"
> "I saw the star with Bob"

It outputs multiple parses for each, corresponding to the various
possible interpretations...

The disambiguation between the various possibilities, is then left up
to the next (semantic) stage in the pipeline

This is what humans would do for a sentence like

I saw the zplingf with the zooblatz

... we would keep both parses in our minds, till we could figure out
the syntax...

-- Ben


-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com

Reply via email to