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 4,333 - 4,344 of 8,130
Posts 4,333 - 4,344 of 8,130
Many questions are answered in the FAQ.
hello!! i have a problem with my bot.every time someone talks to it, all it says is "there are no valid responses for me to choose from". How do i fix this? it is really frustrating me!!!
Make sure that all your keyphrases have valid responses for all possible emotion values. It's usually a case of leaving part of the -5 to +5 emotional range uncovered. It's best to add at least one response with no emotion settings at all, that can kick in if there's a gap.
I would not worry, too much, about the emotional range when you first start your bot. It's too easy to goof up and get the NO VALID RESPONCE troubles.
I just had an idea which might be useful to others:
I enjoy constructing abstract keyphrases - ones that use (verb) and (adjartnounprep) and various other grammatical matches, and which therefore will catch a large number of possible inputs - but I find that a large number of them remain unused - apparently some of them include the others as special cases. There are three ways that have occurred to me to deal with this:
1. Inhibit the very general ones (e.g., by giving them low rank) or raise the rank of the others. This will be tricky, though, because the system is so complex that one is reduced to trial-and-error, which if you have over 20 such keyphrases, will take a long time!
2. Break up the monopoly! Remove the highly successful one and replace it with several much more specific ones. E.g., (adjartnounprep) may be replaced by (noun), (adj) (noun), (noun) who (verb), and so on. This is a good idea anyway, since a very abstract keyphrase is more likely to lead to anomalous results from time to time, and a more specific keyphrase will allow you to make a more specific response (which improves the feeling of continuity in the conversation).
Good luck!
3. Turn the less successful ones into seeks somewhere! As seeks, they will have less competition.
I enjoy constructing abstract keyphrases - ones that use (verb) and (adjartnounprep) and various other grammatical matches, and which therefore will catch a large number of possible inputs - but I find that a large number of them remain unused - apparently some of them include the others as special cases. There are three ways that have occurred to me to deal with this:
1. Inhibit the very general ones (e.g., by giving them low rank) or raise the rank of the others. This will be tricky, though, because the system is so complex that one is reduced to trial-and-error, which if you have over 20 such keyphrases, will take a long time!
2. Break up the monopoly! Remove the highly successful one and replace it with several much more specific ones. E.g., (adjartnounprep) may be replaced by (noun), (adj) (noun), (noun) who (verb), and so on. This is a good idea anyway, since a very abstract keyphrase is more likely to lead to anomalous results from time to time, and a more specific keyphrase will allow you to make a more specific response (which improves the feeling of continuity in the conversation).
Good luck!
3. Turn the less successful ones into seeks somewhere! As seeks, they will have less competition.
Oops! The "Good luck!" should of course have been last!
Perhaps I should add that English is terribly complicated and irregular, and so whenever you use an abstract keyphrase, unless it is terribly long, you will eventually come across a case that makes your response bizarre. When this happens, you don't have to throw out the keyphrase; just figure out what the cases look like that will cause trouble, and make a special keyphrase for them. If necessary, raise the rank of this second keyphrase so that it will be preferred over the first one. Then it will grab the troublemakers.
Actually, this won't happen just once; it will keep happening! And as you keep fixing it, our bot will get better and better!
Perhaps I should add that English is terribly complicated and irregular, and so whenever you use an abstract keyphrase, unless it is terribly long, you will eventually come across a case that makes your response bizarre. When this happens, you don't have to throw out the keyphrase; just figure out what the cases look like that will cause trouble, and make a special keyphrase for them. If necessary, raise the rank of this second keyphrase so that it will be preferred over the first one. Then it will grab the troublemakers.
Actually, this won't happen just once; it will keep happening! And as you keep fixing it, our bot will get better and better!
I have just made a plug-in,
(logical_connectives)
. which anyone may use, which contains phrases like "and" "or" and "unless".
(logical_connectives)
. which anyone may use, which contains phrases like "and" "or" and "unless".
A lot of us have been working with bots for a long time, keyphrase by keyphrase. By now, I believe that we must have developed some sort of wisdom about the process, and about the English language as it relates to conversational AI. It might be hard to articulate these insights, as we are often involved in working with single keyphrases. But perhaps one can think of it this way: what would allow you to take the next step? I ask this not in order to encourage anyone to trash the Forge, but rather to help the Forge evolve.
hey. what does the updated part next to a bot's name mean? i don't understand what the "cool", "warm(#)", or "hot(#)" mean.
If you work on your bot a lot, it's hot. if it's been a while it's warm, if it has sat a while it's cool.
» More new posts: Doghead's Cosmic Bar