[plt-scheme] Natural Language parsing in CS1

From: Eli Barzilay (eli at barzilay.org)
Date: Mon Jun 1 15:22:23 EDT 2009

On Jun  1, YC wrote:
> The AI books that I had read before took the parsing approach,

Yes, this was a typical example in generic AI books.  Probably dating
back to when people thought that this would be relatively easy.  (I
don't remember where exactly, but there's a story from the early days
of CS where parsing natural language was given as a homework

> so I am not familiar with the alternative - would you happen to have
> links or googling phrases for finding the alternatives?

Modern NL parsers are all statistical tools.  They usually begin with
a large corpus of human-parsed text, and learn how to parse more data.
And there's significant work in making the learning process effective,
and generalize it.  The shift to this direction happened roughly
during the 90s, and now I don't think that there are any practical
tools that are using the symbolic approach.  The bottom line is that
the statistical shift makes this a very different field from what is
presented in common AI books -- one thing you'll see is that there is
much more "serious math" involved.

As for a google phrase, a simple "natural language parsing" should
work.  You can also try google scholar if you're really interested
(but in that case pay attention to the dates, since post 2000 papers
will be very different than pre-1990).

[BTW, just to clarify -- I'm not an expert in NLP by any measure, I
just happen to live with one so I have extensive second-hand
knowledge, whatever that may be worth...]

          ((lambda (x) (x x)) (lambda (x) (x x)))          Eli Barzilay:
                  http://www.barzilay.org/                 Maze is Life!

Posted on the users mailing list.