A conversational system with flexibility will require an
intelligent control system. The usual linear system x[k+1] := A*x[k]+B*i[k]
can be fashioned into the usual control system, but that is less flexible than
I would like the conversational system to be.
Here is an adaptation of linear systems to intelligent
systems by adding a value judgment (VJ) subsystem to it. The whole progression
from standard linear system to value based system is shown at:
https://en.wikipedia.org/wiki/Real-time_Control_System
The so called RCS-4 version, the recommended intelligent
control system, is described on that link, as summarized in overview here:
Value state-variables define
what goals are important and what objects or regions should be attended to,
attacked, defended, assisted, or otherwise acted upon. Value judgments, or
evaluation functions, are an essential part of any form of planning or
learning. The application of value judgments to intelligent control systems has
been addressed by George Pugh.[16]
The structure and function of VJ modules are developed more completely
developed in Albus (1991).[2][17]
The Pugh reference is to RCS-4 concepts, while there is
another Albus reference which includes figure 1 as the architecture overview
below:
That architecture diagram is from:
"Outline for a Theory of Intelligence", which
is in free PDF below:
ftp://calhau.dca.fee.unicamp.br/pub/docs/ia005/Albus-outline.pdf
The value judgment (VJ) module appears to steer the logic
behind the Planning and Execution side of the diagram, at the
whims of the World Model Database, while the old standby Situation
Assessment side figures out what can possibly be thought, while the VJ
module appears to figure out which thoughts work best of those available, and
the planning and execution model decides what to focus on, schedule and do.
A paper titled "A Value Driven System for Autonomous
Information Gathering" is here:
http://rbr.cs.umass.edu/papers/GZjiis00.pdf
Here is a fast summary of
that paper's content:
And finally, there needs
to be a script of textual utterances, with patterns to be matched against variable
bindings. Each node in the DAG should also have a slot for some function
capable of estimating the value of each utterance to each goal, and the duration
and cost of each utterance to utter and process.
That is where the DAG
comes in, IMHO. Every statement that is matched would have follow on
questions to ask, together with a new set of expected patterns to be matched.
The highest value, lowest cost question designed to elicit an answer previously
unknown and undeducable from the current world model would be one way to choose
the next question.
But variation helps make
the utterances more interesting. So each node in the conversation DAG should
be a possible child of a branching node which has both possible utterances as
children, perhaps many more nodes, each with a possible next conversation move.
Reviews of the discourse
representation systems (DRS), especially Kamp's should help interested readers (like
myself) to wrap some meat around those bones, so here is a reference to Kamp's
work I found in PDF form:
http://www.ims.uni-stuttgart.de/institut/mitarbeiter/uwe/Papers/DRT.pdf
Does anyone have
references to a text generation paper they especially like?
Sincerely,
Rich
Cooper,
Rich Cooper,
Chief Technology Officer,
MetaSemantics Corporation
MetaSemantics AT EnglishLogicKernel DOT com
( 9 4 9 ) 5 2 5-5 7 1 2
http://www.EnglishLogicKernel.com
-----Original Message-----
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx
[mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of John F Sowa
Sent: Tuesday, July 14, 2015 8:35 PM
To: ontolog-forum@xxxxxxxxxxxxxxxx
Subject: Re: [ontolog-forum] Ontology based conversational interfaces
Rich and Tom,
1983 was one of the "boom times" in the
boom-and-bust cycle of AI.
That's when AI researchers were getting LISP machines and
high-end workstations. There was a lot of optimism about getting truly
intelligent systems. The Cyc project was started in 1984 with a 10-year
plan to solve all the problems.
RC
> a free pdf about discourse and conversational
analysis:
>
https://abudira.files.wordpress.com/2012/02/discourse-analysis-by-gill
> ian-brown-george-yule.pdf
TJ
> I think it's definitely worth a read, although,
being published in
> 1983, most of its value probably lies in documenting
the history of
> discourse analysis...
That book does a good job of surveying the complex issues
about the semantics of natural language and the many, many ways that language
is related to context, speakers, presuppositions, etc.
And they also show the huge number of reasons why we
still do not have computer systems today that can understand natural language.
For just one of the many reasons why formal systems for
NLP have failed, look at page 80 of that book (if you're using the Adobe
reader, it's p. 47):
> In this approach, each participant in a discourse
has a presupposition
> pool and his pool is added to as the discourse
proceeds. Each
> participant also behaves as if there exists only one
presupposition
> pool shared by all participants in the
discourse. Venneman emphasizes
> that this is true in 'a normal, honest discourse'.
The last line is a typical method for dismissing all the
hard parts.
The authors of the book recognize and discuss the many
complex issues involved in that assumption. Unfortunately, what Venneman
calls "a normal, honest discourse" rarely, if ever, exists -- I don't
believe that the terms 'normal' or 'honest' are appropriate.
Unfortunately, the boom years of the 1980s were followed
by a typical bust, when people realized that language understanding is much
harder than anybody in realized. I often quote Alan Perlis:
"A year spent working in artificial intelligence is
enough to make one believe in God."
Those issues are the theme of a talk I presented last
year on "Why has AI failed? And how can it succeed?"
http://www.jfsowa.com/talks/micai.pdf
John
_________________________________________________________________
Message Archives:
http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr:
http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/ Community
Wiki: http://ontolog.cim3.net/wiki/ To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J