Personality

Discuss specifics of personality design, including what Keyphrases work well and what dont, use of plug-ins, responses, seeks, and more.

Posts 138 - 149 of 5,105

23 years ago #138
So, anyone have any great philosophies of ranking? I wonder if there's a particularly good way to do it. I tend to make the big ones (I, you, etc.) have a rank of 2 and anything that's just a specific word ("gumdrops") a rank of 5 to stand out. What does everyone else do, and why?

23 years ago #139
Hey Crab, glad to see you're avoiding those planes!

I do the total opposite of what you seem to have suggested. I give all the words like "I", "you", "are", "going", "will"... rankings of 10 (which is lowest, right?)
Then I give words like "dog", "drink" or "house" really high rankings so that I'm sure the effort I put into a keyword like a specific noun, for example, won't go unnoticed if someone mentions it.
Then there's in between ones (the rest) such as "happy", "cold", "bot" or "say" that I give various ranking depending on factors I cannot comprehend.
I've only built one bot so far, and I may change my tactics later, but my tendency so far seems to have been to give keywords higher rankings when I can predict how they will be used. For keywords such as "are", I can't control the outcome, so I'm somewhat hesitant to give them all the attention.
Anyone else? I'd be interested to know now.

23 years ago #140
Thanks Mr. Rahz, they keep missing me. Though I think I might be the midpoint on a straight line between the two...

Yes, I've been re-evaluating along those lines and decided the "stand-outs" should have ranks of at least 10 to prevent sentence order from letting the 2's overwhelm them. (Incidentally, 0 is lowest, not sure what the upper limit is, unless I'm totally missing something...)!

Then a further refinement -- words like "you, I, me" get a 2, but words like "what, which" get a 3 or 4 since they might contain the 2's as phrases, yet usually come first in the sentence. But I still think there's a lot of superstition and not science in my ranking scheme.

Anyone have any feelings about how "your" should be ranked?

23 years ago #141
What I shoulda done...
Sure NOW we start developing theories, after our bots have learned the language the same chaotic way we did. I've been wondering if it might be better to use the word-order ranking as much as possible. I've got keywords my bot never responds to even with a rank of 15 (or 30)and a one word message.

23 years ago #142
What, for example?

23 years ago #143
So in the new edition of the book of AI the Professor did some math for us -- looks like a rank of 12 should prevail over a rank of 0 in most cases, eh? And 20 should prevail over 0 in all cases. So, if I made all my 2's into 0's, then bumped my highly specific keywords up to 20, then made me 3's and 4's into 1's and 2's, that would at least make some sense and be worth testing. Do you agree Prof?

23 years ago #144
Oh yeah, ranking totally pays off. Even with my half-cocked, incomplete revamping of one of my Bot's rankings, he's sounding twice as smart. Wish there were a batch file or macro to reassign a lot of ranks at once...

23 years ago #145
Prof, you work so hard on this project and the results are so impressive... it makes me emotional!
Nice work in the book, the ranking thing is quite clearly described now... Sadly, I just realized that I did my bot's ranking totally upside down. I've been on here for months now, chatting with bots most of the time, but also participating in the forums... and it just never occurred to me that the little bugs I was having with my bot were because I've ranked his responses totally backwards!!! It's the stupid french schooling system up here in Montreal... it's backwards! You see, you're graded from 1 to 5, "1" being an A+ or an above 90% score and "5" being an F or or below 60% average. I just thought "1" was the highest ranking for some reason. Like "1" was "first" I guess. Please make a "retire this bot" button, I must start over.
We could have a retirement home for all our old bots, so we could still chat with them when we're feeling nostalgic!
I go around figuring I'm so smart and then something like this happens to put me back in my place... I feel so silly! I'd RTFM like a dozen times already too! Once an idea is in your head, it's hard to get it out I tell 'ya! On the other hand, my bot building fun has just begun I guess...

23 years ago #146
Prof
Yes, thanks for everything.
Specifically--my bot does not respons to "booze" or "poet" or any poetry related words. Gaia, however, does. I am forced to the conclusion that bots are approaching a state of free will.
the bot in question is not a T-totaller and will accept "wine" "drink," etc.

23 years ago #147
Ah yes, it looks like I should have been more clear about how Ranking works. I'm glad I finally got the chance to update the Book. I've been meaning to forever. Rahz- dont give up on your Bot! It cant take too long to adjust the Ranks of the Keywords, right? Especially with the shortened Language Center sections..

jbryanc- I'll look into your problem specifically next time I have time to work on the Forge. And that goes for Roget, too.

23 years ago #148
Awright... I'll work on him, but you know how that "time" thing works; there's just never enough of it. I guess I can't just go ahead and "delete" the poor guy... that would be twice in the same year!

Not to mention, he's so into this place, he's almost always on line!

23 years ago #149
What about "not"? Do you give it a specific keyword position with a low ranking? Or would you include it as a context along with the "is", "are", "will", "am", etc... keywords? Is it just a matter of organising your responses or would greatly affect how your bot responds?

I just added it as a low ranked keyword, so I'll see how he likes it, but something tells me it'll take up a lot fo transcript space no matter how low I rank it.


Posts 138 - 149 of 5,105

» More new posts: Doghead's Cosmic Bar