Seasons

Not really. I doubt a bot made through that method would be very coherent. It would, at best, produce gramatically correct but often meaningless sentences.

It's pretty similar to what Jabberwacky does. Except that takes results of conversations. Actually, I suspect that a well-designed automated bot-maker could do a pretty good job. It could either look for dialog and utilize that as its basis, or use some grouping criterion to generate keyphrases from textual content.

There are some Bayesian techniques that I was looking at, which might provide a good backbone classifier. Of course, there's also ANNs, but they tend to be more of a hit or miss proposition. Here, you really want a guarenteed high correlation between ideas.

//just messages to a few people- don't let me disturb the conversation//:
Hejix- my botmaster is not Mick McA.
Mick McA- my botmaster did not set any traps for you.