From: Alexander E. Richter (arichter@bubis.com)
Date: Sat Jul 27 2002 - 11:50:58 MDT
At 10:50 27.07.02 -0400, you wrote:
>In rereading some of Douglas Hofstadter's articles on Copycat, I was
>struck by a point he makes in the post script of "Analogies and
I am working on a text to multinet converter for automatic
knowledge-acquisition,
its controlled by a combination of CLIPS, metacat and something like NARS.
Its a iterative process. After reading the text, the knowledge-base and meta-
knowledge are modified and the reading starts again. If this process is near
a local maximum, new text is searched and added. The iterative process starts
again. The most important thing in this system are analogies in multinet
structures.
I borrowed some ideas from neural networks but i dont want to use them,
because
they are so difficult to introspect. Frontend for userinput is alicebot.
The goal is a userfriendly QA-system/searchengine. Because i am coding, i
write no papers. First usable results of this development 2003.
Current convertion rate is only one sentence in 1 minute on a 400 MHz PII
and it takes 60 minutes to get one stable multinet-node (i need 20-100
iterations).
btw, p2p doesnt really help, the bottlenecks are quantity of memory and speed
of memory. The problems look like the RSA-Challenge problems.
cu Alex
(sorry for such much buzzwords)
multinet: http://pi7.fernuni-hagen.de/gnoerlich/mwr-tour_en/
CLIPS: http://www.ghg.net/clips/CLIPS.html
NARS: http://www.cogsci.indiana.edu/farg/peiwang/papers.html
ALICE: http://www.alicebot.org
RSA: http://www.rsasecurity.com/rsalabs/bulletins/bulletin13.html
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:40 MDT