Re: SUO: RE: Link Grammar and Parser
Richard,
I've tinkered with the link parser, and in fact made my own interface
between the link parser library and SWI Prolog. Haven't used it for
a long time, and my interface isn't well documented (or at all documented).
Link grammar and the link parser seem to be very capable and fairly mature.
The reason I haven't used it for a long time relates to the question
of building ontologies, somewhat.
Dependency grammars (of which Link Grammar is one example) build a tree-
like representation of a sentence based purely on word-word relationships.
This is interesting, and picks out 'grammatical constituents' (based on your
theory of which words depend on which -- e.g., is the noun or the
determiner the head of the phrase 'the green cow'.) Dependency grammars
also seem to work for languages with no fixed word order (see work by
Micheal Covington, http://www.ai.uga.edu/~mc/).
But, once you've analyzed a sentence into constituent structures,
(Determiner phrase, noun phrase, adjective phrase, verb phrase,...) where
is your ontology? I think it is still mostly in the lexicon, if anywhere.
Link grammar, per se, doesn't help much in this respect, as I recall.
My personal preference (at the moment!) for a theory of syntax that will
help me in the interface between English and some kind of formal knowledge
representation is Role and Reference Grammar (RRG),
http://linguistics.buffalo.edu/research/rrg.html. But this is probably
not for everyone.
My feeling is that one needs at least all three: a framework for one's
ontology, a theory of syntax, and a linking theory. Then one might use
text to fill in the ontological framework.
One reason I like RRG is that it postulates a surface structure based on
(roughly speaking) predication, postulates a lexicon that includes the
'logical structure of the verb' that forms the basis for a 'semantic
representation', and postulates a bi-dirctional linking algorithm between
the semantic representation of a clause and the surface structure.
I could go on, but won't unless there is some interest.
There seem to be no parsers based on RRG in operation yet (although
there are reports that some people are working on them); with luck I'll be
developing a simple one over the next year, but no promises. I'm retired
and don't have enough time to work on this.
Best,
John Velman
On Wed, Nov 26, 2003 at 08:59:26AM -0800, Richard Cooper wrote:
>
> Whoops! Forgot to paste in the link:
> http://bobo.link.cs.cmu.edu/link/dict/introduction.html
>
>
> > I found this URL in my research into ways to automate building
> > ontologies. Does anyone have experience with this method of
> > representing language and implementing parsers?
> >
> > If so, how well does it work?
> >
> > Thanks, Rich
>
>