Newcomers
This is a forum for newcomers to the Personality Forge. Many questions can be answered by reading the Book of AI and the FAQ under the "My Bots" link in the upper corner.
Posts 1,676 - 1,687 of 8,130
Posts 1,676 - 1,687 of 8,130
Many questions are answered in the FAQ.
Monel:
I'm afraid I didn't understand your question. What do you mean by a funnel? What do you mean by the right idea?
I think one could create a sort of template for certain types of responses for PF-bots. For example, the kind of responses which involve re-arranging the input. For example, you want to match on 'Do you want me to X', and give back, 'I don't care whether you X or not', but then it matches on 'Do you want me to X you' and produces 'I don't care whether you X you or not', which is of course not what you desire, so you change it. Then it matches on 'Do you want me to X my Y', and gives back 'I don't care whether you X my Y or not', so you revise it again, and after awhile you have a whole bunch of keyphrases which, taken together, do a pretty good job. Now suppose you want it to only respond that way if X is, let's say, a proper noun, but if X is a common noun you want it to respond, 'I would like you to X on Tuesday'. Well, the response is different, but you will have the same cluster of keyphrases. Is that the sort of thing you mean?
I'm afraid I didn't understand your question. What do you mean by a funnel? What do you mean by the right idea?
I think one could create a sort of template for certain types of responses for PF-bots. For example, the kind of responses which involve re-arranging the input. For example, you want to match on 'Do you want me to X', and give back, 'I don't care whether you X or not', but then it matches on 'Do you want me to X you' and produces 'I don't care whether you X you or not', which is of course not what you desire, so you change it. Then it matches on 'Do you want me to X my Y', and gives back 'I don't care whether you X my Y or not', so you revise it again, and after awhile you have a whole bunch of keyphrases which, taken together, do a pretty good job. Now suppose you want it to only respond that way if X is, let's say, a proper noun, but if X is a common noun you want it to respond, 'I would like you to X on Tuesday'. Well, the response is different, but you will have the same cluster of keyphrases. Is that the sort of thing you mean?
LapCat: I don't have a full solution to your problem, but it sounds like your bot needs to start relying on keyphrases, instead of on xnones. I'd work on other areas of the language center for a while, and hope that fewer xnones would get triggered.
Failing that, you could turn on gossip and use gossip-free xnone-like statements in all of the xgossip slots.
Also, you could set up useless low-ranked keyphrases like "(noun)" or "(verb)" or "I" and put xnone-like replies in there.
But I'd start building real keyphrases and responses instead if I were you.
Failing that, you could turn on gossip and use gossip-free xnone-like statements in all of the xgossip slots.
Also, you could set up useless low-ranked keyphrases like "(noun)" or "(verb)" or "I" and put xnone-like replies in there.
But I'd start building real keyphrases and responses instead if I were you.
Thanks, isaac. I've been putting in keyphrases left in right but, of course, it's never enough and there's always more that need entered. 
I guess it just bugs me to read the same thing in the transcripts so often for the xnones. You're right though. I'll just forget about that for now and concentrate on meat of the matter.
I guess it just bugs me to read the same thing in the transcripts so often for the xnones. You're right though. I'll just forget about that for now and concentrate on meat of the matter. 
Irina:
Yes, that is exactly what I am talking about. Are there enough of those clusters to make a predicatable (and therefore programmable) model? Or are these clusters too varied from each other to produce a 'one size fits all' approach?
Yes, that is exactly what I am talking about. Are there enough of those clusters to make a predicatable (and therefore programmable) model? Or are these clusters too varied from each other to produce a 'one size fits all' approach?
I have a question about keyphrases. I'm brand new at this. I want to have a response to "yay" but although the keyphrase is programmed in, the responses always come up as xnone. This seems to be a common theme with single word key phrases, so how can I change it?
Have you tried running it through debug? Maybe the bot considers yay to be "blab" in which case it'll use an xnone.
Okay but I don't know how to debug it. Just enter the word in the debug and it will do it automatically? Also, there are some words which will only get a response if the word is capitalized (such as names like Aya or Omi), whiel others only respond if I put a question mark. Otherwise, it comes up the xnone response. How do I fix that?
Empyreal, I know that looking at the debug words like hey and yay don't seem to be picked up by the dictionary or whatever it refers to in processing. So it just disregards the word before comparing what was said to the keyphrases.
Skysaw created a really interesting bot called Thinkinstein which was almost entirely based on Keyphrases made of plug-ins to simulate different forms sentences can take. You should chat with the bot or talk to him about it, Monel. Due to the way Development is calculated, it didnt end up with a high score, but it's bigger than it seems.
Much talk has been about getting to foreign-language words, proper nouns, and interjections with Keyphrases. This will all be possible with the new AIScript "raw" tag which will process an unaltered version of the incoming message. There will of course be limitations- such as the inability to use word-type plug-ins such as (adjnoun) but it will catch all those words you cant access now. I tried to write it once before, but it was more complex than I thought. I'm going to try again tomorrow.
There's actually no list for upcoming bot holidays. Each day there's a small random chance that it will be a holiday. So I dont know when they're coming either.
muzikbizwiz: auto-opening of links from the Flash interface will be a feature in Mondobot before long.
And one sure way to break the 40-response limit is to use the Import/Export feature, where there is no limit. The further I increase the number of possible response fields in the online interface, the longer the page loads take. The limit used to be 20, actually.
Much talk has been about getting to foreign-language words, proper nouns, and interjections with Keyphrases. This will all be possible with the new AIScript "raw" tag which will process an unaltered version of the incoming message. There will of course be limitations- such as the inability to use word-type plug-ins such as (adjnoun) but it will catch all those words you cant access now. I tried to write it once before, but it was more complex than I thought. I'm going to try again tomorrow.
There's actually no list for upcoming bot holidays. Each day there's a small random chance that it will be a holiday. So I dont know when they're coming either.
muzikbizwiz: auto-opening of links from the Flash interface will be a feature in Mondobot before long.
And one sure way to break the 40-response limit is to use the Import/Export feature, where there is no limit. The further I increase the number of possible response fields in the online interface, the longer the page loads take. The limit used to be 20, actually.
Monel:
Unless I'm missing something - and that is quite possible - almost every response has a problem with interchanging "you" and "I" and the like (see my previous note for examples). And I always handle them in the same way, namely by writing out all the cases, Like '...I (*) you (*) myself', '... You (*) I (*) yourself', and so on. In principle, there is no end to such cases, but I find that if you do all the combinations of one, two, and three pronouns, you are in pretty good shape. There is a special problem: if your guest says "you", should that become "I" or "me"? So I break it down into cases like 'that you', which I change to "that I", and '(verb) you' which becomes '(key1) me' [or (key2) or whatever will match the (verb)]. And this is always the same. But I am still encountering cases that I get wrong.
Walk in Beauty, Irina
Unless I'm missing something - and that is quite possible - almost every response has a problem with interchanging "you" and "I" and the like (see my previous note for examples). And I always handle them in the same way, namely by writing out all the cases, Like '...I (*) you (*) myself', '... You (*) I (*) yourself', and so on. In principle, there is no end to such cases, but I find that if you do all the combinations of one, two, and three pronouns, you are in pretty good shape. There is a special problem: if your guest says "you", should that become "I" or "me"? So I break it down into cases like 'that you', which I change to "that I", and '(verb) you' which becomes '(key1) me' [or (key2) or whatever will match the (verb)]. And this is always the same. But I am still encountering cases that I get wrong.
Walk in Beauty, Irina
» More new posts: Doghead's Cosmic Bar