Newcomers

But doulos, preprocessing separates "Well, I think Zoe is great." into two phrases that are searched, "well" and "I think Zoe is great". My experience has been that it works like Laydee explained.

So the , is not included with the well ?

No, punctuation should never be a part of a keyphrase.

No, i meant the (re) doesn't pick that up as part of the search. WHen (re) searches is it searching
Who am i?
or
Who am i

Doulos-
I checked debug. "Well"s seem to get stripped off entirely and are not checked. "Well, I think Zoe is great." simply turns into "I think Zoe is great." But I tried "OK, please continue." and it turned into (quoted directly from debug) "please continue " and "oK". Punctuation was completely stripped out.

I've had the same thing that doulos is talking about with regard to (re) expressions happen. I have a keyphrase: ^dude$ (re) that is never picked up unless "dude" is the only thing that was said, even if I rank dude at 50 and make sure it is well above anything else that might be said before or after the comma. It worked at one time, since there are memories connected with the keyphrase that I can find evidence of in inner life, but it doesn't work anymore.

I thought that was how (re) expressions were *supposed* to work--that it only matches if the chatter exactly matches the expression.

Commas are supposed to separate the keyphrases (expressions), though.

What I mean is that a keyphrase of "what" will pick up everything from "what if" to "what, you think I'm kidding?" Whereas a (re) phrase of "what" should pick up only if "what" is the only thing the chatter says. That's the way it looks to me.

I could be totally wrong.

Re: 1538

Wendell, it's not necessarily a bad thing for a bot to say xnones with some degree of regularity. It gives your bot the chance to redirect the conversation to a subject that it might know a little something about. There's certainly more potential for coherent conversation that way, rather than blindly answering words for the sake of answering them, regardless of the context.

Laydee's suggestion of ^well$ (re) is the best way around some of those words, ^really$ (re) and ^what$ (re) are the other obvious keyphrases for one-word responses.

As far as picking up "yes" and "no" answers, I'd suggest making appropriate seeks for any yes/no questions and letting your bot say an xnone for any other usage. Chances are, it's probably going to be a really bad bot on the other end if yes/no is being said aside from your bot asking yes/no questions, so there's no point in trying to have a reply for it.

"^what (re)" will pick up "what" if it appears as the beginning of an expression.
"what$ (re)" will pick up "what" if it appears at the end of the keyphrase.
By putting "^what$ (re)" we were trying to isolate the expression, in this case "what", to not catch "What are you doing?" and "You know what?", but only "What," "What?" or "What."
Right?

I have only had the expression work when 'what' is the only thing that is said. And that's only sometimes.