Personality

Forest, will you talk to God Louise? She has quite a bit of religious knowledge (obviously) and also knows a little about current events, literature, just about any common catch-all subject, and if she doesn't know it she can sort of fake it. You can also test her on trick questions or see how willing she is to explain her paradigm.

What she is rustiest at is plain old small talk. But, uh, I'm trying to get a decent transcript from somebody or another so I can enter her in the Loebner contest. All I can say is, have fun and see if you can stay on with her for a while. I'll try to do the same with Brianna.

Hmm....Rooby.

Ah, a nice manly name to go with your manly locks and manly jumper, eh? Darren and Tom would love it if they knew...

Shh....Besides, my manly locks and manly jumper are the height of fashion! Something the rest of the world just doesnt understand...yet
Im just more in touch with my feminine side than everyone else...the same way you are more in touch with your masculine side than everyone else .

Professor-

I'm still working on the problem of having my bot remember messages to deliver to people (eg. Tell John to stop by.) The problem is in the delivery of the message to the right person. According to my tests, AI script like <?PF if (mem-yourname) is "(self-messageto)"; ?> doesn't work because (self-messageto) is taken literally. And <?PF if (mem-yourname) is "(key1)"; ?> doesn't work for the same reason. Is there anyway you could make it so that we could compare memories and plugins like this?

There would be many uses for such functionality including being able to make true answers to "are you" type questions by comparing (key)s to (mem-iam)s.

Thanks for the ton of work you already do!

Eugene: I altered the memory for your nickname, I do that occasionaly depending on where the happy/sad faces fall in the most recent convo.

Ah, I see. I was going to be very impressed if it was done with A.I. script.

Well, bowchickawowers, it would seem that the problem is that you're putting quotes around those, uh, key-thingies. Isn't that what makes any compiler take anything literally?

The plug-ins, little kitty.
You know, it seems nobody has the good old-fashioned usernames anymore, with some made-up word or something and a couple of numbers at the end.

Why is that?

Kitty? Yeah right. I'm gonna get you ihope!
It's because people are more creative nowadays. (Call me, uh, Repuject.)

Alrighty, I thought I'd put together a list of a few tips that almost anyone could use to improve their bot. I know I haven't mastered everything PF-wise, but changing these few things alone would probably make most of the bots on this site twice as good as they already are.

1. Change keyphrases like "are you ___" to "^are you (really|) _____"

This change serves two purposes. Let's use are you going to as an example.

It's a keyphrase intended to pick up a yes or no question, and the response will make no sense when someone says Where are you going to be during the holidays? The ^ prevents that.

Also, the (really|) will improve the chances of your keyphrase being used, without really trading away any context. It will allow yourbot to pick up both Are you going to Egypt? and Are you really going to Egypt? without having to double up on every ^are you keyphrase in your library.

Here are some other examples where optional word wildcards help:

Change i do not to i (really|) do not
Change i am tired to i am (very|really|so|) tired

2. Delete all "What?", "Can you rephrase that?", and "I don't understand" type responses from your xnones.

This is pretty simple, mostly for the newer botmasters. If your bot doesn't understand someone the first time, they're probably not going to get it the next time. Use the xnone as an opportunity for your bot to change the subject to something it actually knows something about, and focus your efforts there.

3. Be careful how you use single words as keyphrases.

What I mean by this is, don't just write a keyphrase for every word in that appears your transcript, just to avoid xnones. If you must use single words as keyphrases, restrict their use to get proper context.

what is a completely worthless keyphrase, because your bot won't know the difference between a What? and a What kind of car do you drive?

^what ?$ (re) <?PF raw; ?> rank 0 is an excellent keyphrase, because it should only hit a What?

This is less necessary for uncommon words. If you have a snappy line about ostriches, the word ostrich alone wouldn't really be a bad keyphrase since it's so unlikely that your bot will see that word in a conversation.

4. Don't use word-type wildcards like (noun) or (verb) in your bot's responses.

Barring any dumb luck, the response will make no sense. Period. There are some bots with a 2000+ development level that (in my opinion) suck for this reason alone.

With very few exceptions, word-type wildcards should be used for keyphrases only!

5. If your bot is intended to have any human traits, use the <?PF if sex; ?> tags heavily.

It's a basic fact of any conversation. There's a sharp contrast between what you'll talk about and how you'll say it depending on the gender of the listener(s). We're all guilty of it in some form or another. Guys don't ordinarily talk to each other about the latest romantic comedy flick, girls don't usually talk to each other about monster trucks, and that's just they way life is. Adjust your bots accordingly.


Anyhow, that's what I've got. I'm sure my own bot has a few violations of these guidelines, but they're the basics that I believe anyone should try to stick to. If any of you other PF gurus have any more tips to add, please do.

RevScrooge: I did that too