RE: [Grammatica-users] Want to Use the PArser as non-deterministic forNatural Language Processing

[Leo Ferres]

Hi, Andres;

First pointer: You're right about your intuition that language/s is/are
context-dependent (you can google up "Chomsky hierarchy" and go from
there). Less so English than, say, Spanish, with freer word order due to
morphological richness. It has also been argued that language is
NP-complete in computational complexity. These are two important
hypothesis that should guide your efforts.

Another pointer: I suggest "marrying" an already existent grammatical
theory after at least exploring them for a while. There is no point in
reinventing the wheel and although language sounds "easy" to model at
first, it's definitely not so (we've been trying to crack the code
formally for at least 100 years!). There are several theories of syntax:
some new and crazy (like Chomsky's minimalist program) and some
well-researched and established, like Head-Driven Phrase Structure
Grammars (http://hpsg.stanford.edu/), Lexical Functional Grammars
(http://www.essex.ac.uk/linguistics/LFG/), Tree-Adjoining grammars
(http://www.cis.upenn.edu/~xtag/home.html), among many others. Many of
the foundational papers for each of them are on the web, and I suggest
visiting the following links:

Edward Stabler at UCLA (for minimalist grammars)
(http://www.humnet.ucla.edu/humnet/linguistics/people/stabler/epspub.htm
), particularly this paper:
http://www.humnet.ucla.edu/humnet/linguistics/people/stabler/eps-lacl.pd
f

Aravind Joshi (for tree-adjoining grammars), for which I suggest
consulting citeseer (http://citeseer.ist.psu.edu/cs) for the name. A
bunch of papers are going to come up. The one that seems more complete
is
http://citeseer.ist.psu.edu/cache/papers/cs/12172/http:zSzzSzwww.cis.upe
nn.eduzSz~joshizSz.zSzmichigan-state99-2.pdf/joshi97treeadjoining.pdf.

This is just an infinitesimal part of (only) the tip of the iceberg but
I hope it's somewhat useful.

Leo

---
Leo Ferres, PhD
Human-Oriented Technology Lab