Hae you checked out Eliza? It's a classic example of a pseudo-AI that manages to hold a coherent conversation with the user by recognizing key phrases and asking questions in the manner of a psychoanalyst.
It's actually pretty simple (we implemented a Eliza in scheme as a lab assignment at school), but very effective. I read somewhere that some people preferred talking to Eliza rather than a human psychoanalyst, because the felt the program made them more comfortable.
There is of course a perl implementation called Chatbot::Eliza. I haven't played with it myself but it looks like it could be useful for you.
Cheers,
--Moodster | [reply] [Watch: Dir/Any] |
On an interesting side note, Eliza has been implemented in BASIC - but not just any BASIC, rather, the BASIC that has itself been written in Parrot assembler. Last I checked she was playing secretary for some of the Parrot mailing lists :-)
Cheers,
Erik
| [reply] [Watch: Dir/Any] |
| [reply] [Watch: Dir/Any] |
I would suggest, other thant the most needed TOP-DOWN approach,
A quick analysis of what topics 'Talbot' (talking bot... Oh my am I so creative today :p ) should be able to handle.
For example, you could put some links:
Talk to talbot about weather
Talk about music
Talk about sex
Each link should load a single interaction database and give a certain amount of key words to use in the questions, to restrict the field of action of the AI.
Then you should figure what kind of response generator you like:
Chance- based, on reasonable combinations of THE QUESTION WORDS, and combinations of ANSWER TEMPLATES with a database for WORDS in the answer!
Pattern- based, pattern matching the question (for syntax) and offering a TEMPLATE ANSWER, with words from the database.
Hope this helped... but I'm sure it did :) | [reply] [Watch: Dir/Any] |