Newsgroups: comp.ai.nat-lang,comp.ai,comp.ai.neural-nets,comp.ai.shells,comp.ai.fuzzy
Path: cantaloupe.srv.cs.cmu.edu!europa.chnt.gtegsc.com!howland.reston.ans.net!ix.netcom.com!netcom.com!sujo
From: sujo@netcom.com (Suresh K Jois)
Subject: NL parsers that learn
Message-ID: <sujoDE35rH.7rx@netcom.com>
Organization: NETCOM On-line Communication Services (408 261-4700 guest)
X-Newsreader: TIN [version 1.2 PL1]
Date: Tue, 29 Aug 1995 18:52:28 GMT
Lines: 37
Sender: sujo@netcom19.netcom.com
Xref: glinda.oz.cs.cmu.edu comp.ai.nat-lang:3786 comp.ai:32955 comp.ai.neural-nets:26559 comp.ai.shells:2664 comp.ai.fuzzy:5531

Hi,

I'm looking for an Natural Lang understanding system that can do the 
following:

* I initially prime the system with some domain specific knowledge
  (for instance airline reservation)
* I throw at it several examples of questions pertaining to the domain,
  and it uses these to refine its knowledge base.
* I use it as a query interface to data in the domain it was trained
  for, and it responds with hi-relevance answers. 
* It has a learning subsystem that can accept feedback of corrections
  to its erroneous responses, and improve error rate.

Are there any compelling reasons NOT to use NL as a frontend to a 
knowledge query system ? I know Natural lang systems have been very 
brittle in the past. But I have the following that I think makes my 
application a little less susceptible:

* I can restrict the user interface such that the user cannot type
  free form NL input, but instead constructs a query from a restriced
  set of "skeletal phrases", with suitable modifiers.

* The knowledge domain is narrow.

I guess what I'm looking for is some kind of domain-specific version of
a story understanding system.

I'd appreciate any help on this, especially pointers to commercial 
systems and survey papers on research/prototype systems.

- Suresh
-- 
---------------------------------------------------------
SURESH K JOIS  CyberSpace Gypsy  Virtual Spaces Architect
sujo@netcom.com    http://www.internet.net/People/Suresh
Traceroute: ->Informix->SiliconGraphics->Sun Microsystems
            ->Home Shopping Network/Internet Subsidiary
---------------------------------------------------------
