<snip/>
> Actually, I've started work on a prototype lattice that
uses WordNet's noun
> classes. So far, I have a first order
Q&A function to move from one node
> in the lattice to
another. I think the same thing is needed for the verbs,
> but
haven't gotten that far.
Doesn' the work on "Formal Concept Analysis "
(FCA) at Karlsruhe build concept lattices something like this?
For
exaple Philipp Cimiano, Steffen Staab, & Julien Tane tested an assumption that
:
"verbs pose strong selectional
restrictions which could be used to build a conceptual hierarchy on the basis of the
inclusion relations between the extensions of the
selectional restrictions of all the verbs. Tthe verbs themselves
provide intensional descriptions for each
concept (in
"Deriving
Concept Hierarchies from Text by Smooth Formal Concept
Analysis"). They have a small concept lattice example in the
paper.
I've also seen discussion of mathematical
structures related to FCA (e.g. concept lattices, Galois connections, closure
structures, attribute implications) that seems to be something worth
considering here if people are building
applications.
Regards,
Gary
Gary, I've not gotten so far along yet as to have objects
with properties that FCA relies on to form classes. So far, I only have
the databases that come with WordNet 2.0, which includes the noun
lattice. I do plan eventually to use selectional preferences on
verb signatures a la Beth Levin's original study, but the noun lattice needs
to be in place first before selectional preferences can be dredged up from
corpora and used to categorize the verbs.
In WordNet, the noun lattice can be the basis for later
adding class properties to the lattice, but I haven't gotten that far
yet.
-----Original Message-----
From:
owner-standard-upper-ontology@listserv.ieee.org on behalf of Rob
Freeman
Sent: Tue 3/22/2005 9:52 PM
To: John F. Sowa
Cc: cassidy@MICRA.COM; West, Matthew R SIPC-OFD/321; SUO WG;
cg@CS.UAH.EDU
Subject: Re: Inconsistent models, mapping,
interoperability, and the SUO
John,
On Tuesday 22 March 2005 23:14, John F. Sowa
wrote:
>
> Before spammers started to play games with
it,
> Google's page-rank algorithm was the best guide
> to
finding useful sites among the trash. Some such
> techniques can
be adapted to any method for storing
> and retrieving data, including
the lattice. You
> can add metadata to each theory that says
who's
> using it, for what purpose, and with what
success.
Seeking meaning in a pattern in the connections of your
lattice, like Google's
page-rank algorithm, which seeks meaning in the
pattern of connections
between pages on the Web.
Yes, that's
nice.
Seeking patterns in use would be one strategy. Another would be
to seek
patterns in response to a query.
Which takes me to Rich's
initiative.
On Tuesday 22 March 2005 23:32, Rich Cooper
wrote:
>
> Actually, I've started work on a prototype lattice
that uses WordNet's noun
> classes. So far, I have a first order
Q&A function to move from one node
> in the lattice to
another. I think the same thing is needed for the verbs,
> but
haven't gotten that far.
Rich. Why use WordNet? Why not use words?
That way you can handle your query
"natively".
As you seek paths
from word-to-word you will travel node-to-node, and describe
a pattern in
your lattice, which can be interpreted meaningfully a bit like
the
patterns of Google's page-rank algorithm, as John
suggests.
-Rob
because WordNet has a
large initial charge of nouns in a class lattice that reflects common usage
of those nouns. To start from scratch, I could extract words and build
my own list, but then I wouldn't have any of the collateral information
about synonymy, antonymy, hyponymy, hypernymy, meronymy and all the
relational database descriptions of those relations to start from.
This approach appears
to provide some initial legs up to stand on so corpus analysis can be used
to flesh out the sparse WordNet conceptual structures.
Since WordNet has no
deductive mechanisms, I've added an FOL rule engine, but won't have any
rules to put into it until I find a source of those rules. So far, I
haven't found any way of automating or kick starting the rule base
though. Cyc doesn't relate its conceptual structures to WordNet
(unless someone out there knows something about how its done). There
are a few rule bases (FrameNet, VerbNet, ...) but nothing as widely accepted
as WordNet so far as I know.
I'm just trying to
provide a basic capability to advance linguistic Q&A technology half a
step.
Thanks for the
comments,
Rich