The source code isn't really that complicated (it's bascially just a
regular expression pattern match and a few "Like" statements in SQL
Server 2005). You can get the basis for this type of code from Eliza
(lots of variations and some
source code in the links there)
Reasoning is a *MUCH* more complicated affair. Cyc is probably the
closest right now using "common sense" rules. For mine I'm hoping to
add hueristics as I go along. The main thing (and you'll find this
is true with all chatterboxes) is getting the corpus of data. That's
why I made mine open to the public to add rules. (The link is in my
post below, I've been shameless enough with the plugs). You, and
everyone else, are of course welcome to add as many rules as you
like. The more rules/responses, the stronger the program becomes..
at least in theory.
Are there really 1700+ members to this group? WOW!!! I didn't
realize how much of an old timer I was. I joined the day after it
started... July of 1999!
Do they make a 7 year pin for nerds? Just askin' ;-)
--- In firstname.lastname@example.org
, Christian M
> Nice little thing. I added a rule to your chatterbox.
> Do you share the source code with anyoneone or do you
> just work on it alone? It would be interesting to see
> it. How does a learning chatterbox work? This is more
> The AI in these chatterboxes is quite simple. They
> have a set of predefined rules. It would be really
> interesting if they would have a the ability to
> reason. This is no small thing I know. The Cyc
> knowledge base tries to do this. There approach is
> ambitious. I wonder how well it works?
> Most work with pattern matching. There's a series of
> rules and
> responses. For example:
> Hello *
> Hi there.
> When a rule matches, one of the responses is choosen
> and the
> program "seems human". In this case the "*" symbol is
> a wildcard. So
> you would get one of the listed responses for "Hello",
> even "Hello, my name is Steve". Theoretically, the
> rules/responses the more lifelike the chatterbox.
> That's why I
> opened mine up to the public-You can get many many
> inputs over the
> internet. (but mine is still in it's infancy)
> The original chatterbox was called "Eliza". It had
> about 50 rules
> that simulated a psychologist. (Basically, it just
> repeated what you
> said with a question). You can probably find it online
> and even the source code. Modern chatterboxes have
> much more
> complicated hueristics, but all use some variation of
> rules/patterns model. Some can even "learn" as they
> Here's a high end chatterbox http://www.alicebot.org
> And mine again *shameless plug*
> Do You Yahoo!?
> Tired of spam? Yahoo! Mail has the best spam protection around