Return to the archive index

RE: rememberance agents

From: Eugene Leitl <>
Date: Mon, 27 Apr 1998 10:48:01 +0400 (MSD)

Jesse Montrose writes:
 > I've had some interest in this area myself, I would enjoy hearing about your
 > progress.
 > 
 > 
 > http://wearables.www.media.mit.edu/projects/wearables/augmented-memory.html
 > I assume you've looked at this one, right?
 > 
 > 
 > Neural Networks can be built to run just fine on a 486, and you might find
 > the processing done in the savant portion of project above to be a good
 > place to start if you want to take that route.
 > 
 > 
 > http://www.signiform.com/tt/htm/talk.htm
 > This project is absolutely fascinating, but has resource requirements that
 > will put it out of your wearable reach.  I recommend looking at it anyway, c
 > source code is available and you may find inspiration.  It also might work
 > out for you to run TT on a more powerful machine, process your text files,
 > and store the results on your wearable for easy lookup.  Here is an excerpt
 > from a sample of its "chatterbot mode" (all these answers come from TT's
 > knowledge base):
 > 
 >       How tall are elephants?
 >      They are extremely tall.
 >       Why?
 >      Elephants are extremely tall, because 3 meters is the height of an
 > elephant.
 >       How big are elephants?
 >      3 meters is the height of an elephant.
 >       Are ants intelligent?
 >      No, they are not intelligent.
 >       Are human beings intelligent?
 >      Yes, humans are in fact intelligent.
 >       What is the most dense atom?
 >      A lawrencium.
 >       What is the least dense atom?
 >      Hydrogen.
 >       What is the circumference of the Earth?
 >      40003236 meters is its circumference.
 >       What is the capital of New Jersey?
 >      Trenton is its capital.
 > 
 > 
 > http://www.dfki.de/lt/registry/
 > If you haven't gone here, it should be on your list (it's a little
 > overwhelming :)
 > 
 > 
 > http://cuiwww.unige.ch/OSG/info/FreeComp/fc/categorie.html
 > This page is worth looking at, excerpt:
 > 
 >      The list catalogues freely available software for language tools,
 >      which includes the following: compilers, compiler generators,
 >      interpreters, and assemblers -- things whose user interface is a
 >      language. Natural language processing tools may also be included.
 > 
 > 
 > http://www.aclweb.org/nlpfaq.txt
 > Last but not least, read the faq :)  (this one is very high-level, offering
 > pointers elsewhere)
 > 
 > > -----Original Message-----
 > > From: R. Paul McCarty [mailto:]
 > > Sent: Sunday, April 26, 1998 6:07 AM
 > > To: 
 > > Subject: rememberance agents
 > >
 > >
 > > Since I can't afford to add wireless networking to my wearable yet, I've
 > > been working on a simple rememberance agent. Basically, I want a plain
 > > text interface to a database (or a NLP interface to an expert system) but
 > > since it has to run on my modest 486, it can't contain things like neural
 > > nets or complex learning algorithms, or text understanding (or maybe it
 > > could, but none of the usual LISP implementations).
 > >
 > > Does anyone know if something like this exists? Has anyone tried to write
 > > something like this? I have a very simple perl script that I wrote that
 > > learns to associate words based on statements I make, for example I can
 > > tell it:
 > >
 > > >my birthday is September 16.
 > >
 > > and then when I ask:
 > >
 > > >when is my birthday?
 > > September 16.
 > >
 > > It's like talking to a five year old, but its pretty neat. I also want to
 > > be able to give it text files to read and learn, but it needs some extra
 > > functionality to do this, like recognizing pronoun references, and being
 > > able to parse more complex statements, etc.
 > >
 > > If anyone has any suggestions, I'd love to hear them.
 > >
 > > Thanks.
 > > -Paul
 > >
 > > R. Paul McCarty / DARS Coordinator /  / x52059
 > > 317 Lattimore Hall, University of Rochester, Rochester, NY 14627
 > > Computers don't make mistakes;what they do,they do on purpose.-Dale/KOTH
 > >
 > >
 > >

Previous Message in Thread | Next Message in Thread

From Wear-Hard Mailing list Archive (WH)
Maintained by R. Paul McCarty

Archive created with babymail