[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Axiom-developer] topic modeling

From: Alfredo Portes
Subject: Re: [Axiom-developer] topic modeling
Date: Sun, 30 Jul 2006 00:36:45 -0400

Classification in fact is an field of Artificial intelligence, more specific a
Natural Language Process sing. More than the current, buzz about "Semantic
Web", the field has had advanced considerably during the last 3 - 5
years period,
because an improvement on various techniques.

As Tim has indicated, IBM has been always on top of this research
involving NLP. This past week was announced how their voice
recognition software does not need user training a breakthrough for
this type of commercial applications.

I am very interested in the NLP area. I consider Tim's idea a very
current and important one. Given the incredible amount of scientific
information on the web, and the amount that is inputted constantly,
the problem of classification and retrieval of this information seems

Having smart systems that can search for this information, classify
and stored them, will be right there with the idea of literate papers
that can be found and added to the system. I will be more interested
in a system that can find the information and create a literate
document from it. As Tim said, this kind of systems perform better
when they are domain specific.



On 7/29/06, root <address@hidden> wrote:
> Classification is a popular subject in AI and especially data mining, I did
> not read the paper but there is a lot different technique. And having an
> article in ZDnet and making a round in the blogospher is not reliable
> indicator of a scientific breakthrough.

I did some work on semantic classification while at IBM Research.
The system was called KROPS, a combination of a rule-based programming
language based on OPS and a semantic network by Eric Mays called KREP.
I showed that there is a dual relationship between semantic nets
and rete-network rules, a unification that allows rules to be expressed
as classification and vice-versa.

KREP uses a special kinds of classification called subsumption which
has provable properties of the classification.

You can read about subsumption here:

KROPS (KREP/OPS) was the basis for IBM's FAME project

KROPS is described in:
 "Integrating rules and inheritance networks in a knowledge-based
  financial marketing consultation system" Proc. HICCS (1988)
  Pages 496-500


Axiom-developer mailing list

reply via email to

[Prev in Thread] Current Thread [Next in Thread]