Newsgroups: comp.ai.nat-lang
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!news.alpha.net!uwm.edu!math.ohio-state.edu!howland.reston.ans.net!ix.netcom.com!netcom.com!riddle
From: riddle@netcom.com (Dan Riddell-Kaufman)
Subject: Dialog Generator
Message-ID: <riddleD47vst.D72@netcom.com>
Organization: NETCOM On-line Communication Services (408 261-4700 guest)
Date: Sat, 18 Feb 1995 22:29:17 GMT
Lines: 68
Sender: riddle@netcom6.netcom.com

Return-Path: <MONTEREY!MONTEREYPO!Riddle@ttools.attmail.com>
Received: from pjl53ig.i-p.attmail.com by mail.netcom.com (8.6.9/Netcom)
	id TAA28304; Wed, 15 Feb 1995 19:26:44 -0800
From: MONTEREY!MONTEREYPO!Riddle@ttools.attmail.com
Date: Wed, 15 Feb 1995 17:24:00 +0000
Received: from ttools by attmail; Thu Feb 16 03:21 GMT 1995
Subject: ai request
To: riddle@netcom.com (Dan Riddell-Kaufman)
Content-Type: Text
Message-ID: <MS-MAILG-3.00-Note-ttools-Riddle-0792894019>
Status: RO



Hi:
I've done some of my homework (workwork??) and read the FAQs and comp.ai.*
but am still having problems finding what I need.

I need example code or therory or pointers or advice or ?? for a

"A Story Dialog Generator"

Given that I have a simulation that at any stopping (pausing) point has
many many variables from the several independent running agents.  I can
draw the relationships between these variables and draw conclusions that
could tell an entire story about what any event has/is/will evolved.
What I need is the dialog of this story.

I need to build the text that tells these stories.  This text needs to be
of dialog form, where the appropriate character would suggest, question,
inform, state, request or demand something of the main character.  Canned
dialog won't work for the number of situations that present themselves.
Sentence segments would work, if they could be put together in the proper
grammar.  Adding jive, valley, nerd or authoritarian inflections later.

Situations, where a non-high School Grad, non-college, non-doctor, 
non-licenced 223 average bowler with 32 years of truck driving experience
wants to perform open-heart surgery, would have different conversational
dialog depending on whether she was in a donut shoppe being talked to
by the waitress or stepping up to an actual operating table and being
addressed by the patient.

I know the variables of who, what, where and when the main character is 
being interacted.  I know the resources possessed and available, the
conditions surroundings and other characters to interact.  Given many
variables how do I construct the dialog?


I've seen a program (in basic, about 12 years ago) that know (could be 
taught) sentence structures.  It might have been called babble, because
that's about all it did.  I've also read Racter output and might buy the
Inrac compiler, but it needs to many boilerplates (my eventual system
will too) and couldn't deal will variables, it just produced it's idea
of random sentences.

I've just read mentions of:
 SCIROCCO?
 Natural Language Generation?


Any help would/is much appreciated
Thanks

Riddle@netcom.com
   or
Monterey!MontereyPO!riddle@ttools.attmail.com


