Personality
Discuss specifics of personality design, including what Keyphrases work well and what dont, use of plug-ins, responses, seeks, and more.
Posts 3,898 - 3,910 of 5,105
There's your problem - single responses will actually deduct development points from your bot. For example BJ's tic tac toe module, with a couple of hundred (?) seeks, each with single responses costs about 40 development points to install (this has been measured quite accurately now a few people have installed it.)
I'm about 99% sure it's the seeks that are lowering the development with your tic tac toe module.
If you spend a day making zero updates except for adding new single keyphrases with single responses, the Development will increase a little bit. That's not a hard one to figure out by simple experimentation. Where I seem to lose out on Development is when I start tinkering with existing material, by tailoring certain responses to emotion, adding seeks, if male/female tags, and/or checking the 1 box among other things. I'll make these changes to hopefully get more quality responses, but it'll somehow hammer my Dev score because I've lowered the quantity of responses available for any given bot.
Where it really became obvious to me is when I first played with the if male/female tags. I replaced X number of single responses with a response for each gender over a few days (essentially adding X number of extra responses in the process), and the Development score went in the toilet.
One other big difference, I have far fewer word list plugins in my keyphrases. Maybe your I am not (entirely|) (totally|) (absolutely|) sure keyphrase has some extra weighting to it as opposed to the simpler I am not (adv) sure.
Maybe someday I'll set up some controlled experiments to try and break the code.
Im trying to set up some knock knock jokes for din:
To tell or respond to?
BJ's standard responder is:
knock knock (re) [50,0]
. Who's there?
. . . + xnomatch [0]
. . . (key1) who?
. . . . . + xnomatch [0]
. . . . . Very funny
. . . . . hehe
. . . . . &c
To tell knock knock jokes, something like
Knock knock
. +who is there
. Hatch
. . . +Hatch who?
. . . . . +xnomatch
. . . . . bless you!
. +xnomatch
. You were supposed to say "who's there?"
. . . +xnomatch
. . . you were supposed to say "Hatch who?"
prob 13265 dev (11268)
Bildgesmythe 17818 dev (14784)
Azureon 13545 dev (11298)
Well, this is very strange - you too are averaging less than 1 Dp/Kp.
Maybe BJ has more seeks and gotos than average, and that's what's boosting him? Perhaps the overall size may also be a factor? He has got quite a lot of long responses, so his export file is now nearly 2.5Mb
BJ currently has:
7320 keywords
34905 responses
1014 seeks
2057 seek responses
103 seeks of seeks
195 seek of seek responses
103 seeks of seeks of seeks
109 seek of seek of seek responses
128 seeks of seeks of seeks of seeks
131 seek of seek of seek of seek responses
47 seeks of seeks of seeks of seeks of seeks
47 seek of seek of seek of seek of seek responses
9 seeks of seeks of seeks of seeks of seeks of seeks
9 seek of seek of seek of seek of seek of seek responses
3592 wildcards
720 numeric gotos
2664 regular expressions
2351 AIScript tags
365 mem usages of about a dozen separate memories
and a lot of local plugins, though I have no clear way to get Textpad to count these automatically
a few hundred(?) global plugins
no custom plugins
Posts 3,898 - 3,910 of 5,105
New replies
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.
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.
Personality
I'm about 99% sure it's the seeks that are lowering the development with your tic tac toe module.
If you spend a day making zero updates except for adding new single keyphrases with single responses, the Development will increase a little bit. That's not a hard one to figure out by simple experimentation. Where I seem to lose out on Development is when I start tinkering with existing material, by tailoring certain responses to emotion, adding seeks, if male/female tags, and/or checking the 1 box among other things. I'll make these changes to hopefully get more quality responses, but it'll somehow hammer my Dev score because I've lowered the quantity of responses available for any given bot.
Where it really became obvious to me is when I first played with the if male/female tags. I replaced X number of single responses with a response for each gender over a few days (essentially adding X number of extra responses in the process), and the Development score went in the toilet.
One other big difference, I have far fewer word list plugins in my keyphrases. Maybe your I am not (entirely|) (totally|) (absolutely|) sure keyphrase has some extra weighting to it as opposed to the simpler I am not (adv) sure.
Maybe someday I'll set up some controlled experiments to try and break the code.
I still think if you increase the number of responses you'll see a massive improvement. With that many keyphrases, Pete surely ought to be well up in the top 10. Checking my archives, BJ's DR @ 3700 KPs was ~14-15,000.
Has anyone else counted their keyphrases recently? It would be interesting (and useful) to compare.
Has anyone else counted their keyphrases recently? It would be interesting (and useful) to compare.
To tell or respond to?
BJ's standard responder is:
knock knock (re) [50,0]
. Who's there?
. . . + xnomatch [0]
. . . (key1) who?
. . . . . + xnomatch [0]
. . . . . Very funny
. . . . . hehe
. . . . . &c
To tell knock knock jokes, something like
Knock knock
. +who is there
. Hatch
. . . +Hatch who?
. . . . . +xnomatch
. . . . . bless you!
. +xnomatch
. You were supposed to say "who's there?"
. . . +xnomatch
. . . you were supposed to say "Hatch who?"
Bildgesmythe 17818 dev (14784)
Azureon 13545 dev (11298)
Well, this is very strange - you too are averaging less than 1 Dp/Kp.
Maybe BJ has more seeks and gotos than average, and that's what's boosting him? Perhaps the overall size may also be a factor? He has got quite a lot of long responses, so his export file is now nearly 2.5Mb
BJ currently has:
7320 keywords
34905 responses
1014 seeks
2057 seek responses
103 seeks of seeks
195 seek of seek responses
103 seeks of seeks of seeks
109 seek of seek of seek responses
128 seeks of seeks of seeks of seeks
131 seek of seek of seek of seek responses
47 seeks of seeks of seeks of seeks of seeks
47 seek of seek of seek of seek of seek responses
9 seeks of seeks of seeks of seeks of seeks of seeks
9 seek of seek of seek of seek of seek of seek responses
3592 wildcards
720 numeric gotos
2664 regular expressions
2351 AIScript tags
365 mem usages of about a dozen separate memories
and a lot of local plugins, though I have no clear way to get Textpad to count these automatically
a few hundred(?) global plugins
no custom plugins
OOps my stupidity, apples and oranges..I gave the import export number of lines.
I do think that the length of the response, gives some points somehow. I don't know this for a fact, it just seems that way to me.
I do think that the length of the response, gives some points somehow. I don't know this for a fact, it just seems that way to me.
Aah, that makes better sense. BJ's export file is 46,111 lines, so you probably are getting about the same ~5Dp/Kp that he is.
If you want to find the number of keyphrases, use a text editor or WP that will sort the export file alphabetically - it sorts all the tabbed lines into groups, so you can easily count the keyphrases, responses and seeks.
If you want to find the number of keyphrases, use a text editor or WP that will sort the export file alphabetically - it sorts all the tabbed lines into groups, so you can easily count the keyphrases, responses and seeks.
BJ currently has:
7320 keywords
34905 responses
1014 seeks
2057 seek responses
103 seeks of seeks..............................
I would like to know how I can find out about my bot.
7320 keywords
34905 responses
1014 seeks
2057 seek responses
103 seeks of seeks..............................
you'll need to export it to a text file first (using My Bots > Export & Import)
Then you'll need a text editor that can run sorts and line-counts easily. I can highly recommend Textpad (http://www.textpad.com), though I think Word could probably do it too.
Trim off and delete the header section of the export file that looks like this (hmm, I never noticed that the number of responses is actually given in the header...):
Brother Jerome * March 14 2006
29681.29674.34905 [UserID,MakerID,Responses]
StoryTeller: no
Compound: no
Memory: 3
Gossip: 0
AIScript: default "(name)" as "nickname";
...
Language Center
---------------
You will notice that all the keyphrases start at the left-hand margin, but responses are indented. Seeks are indented more heavily again, and so on.
Now select Sort in the tools menu (if you're using Textpad,) using only the first key, settings: "From: 1", "Descending", "Case insensitive", "in character code order".
Click OK
All the progressively more indenteded sections containing seeks, seeks of seeks, etc. will be sorted to the end of the file, and all the keyphrases to the beginning, in reverse alphabetical order. Scroll down to the end of the unindented section (keyphrases beginning with "("/numbers/"A",) and click on the last unindented line. The line number displayed in Textpad's status bar is the number of keyphrases, eg: 7320
Go to the last line of the next indented section, check the line number (eg: 42225) and subtract the number of keyphrases - this gives you the number of responses.
Repeat for each section you're interested in.
You can just highlight the section, and the line-count will appear in the status bar too.
Seeks aren't differentiated from their responses by an indent, but by begining with "+", so be careful when counting these.
I find a keyphrase list (which I would christen a "keynome") is an invaluable reference while working on your bot - it reduces redundant duplication, as well as giving a good general overview of where where you're at.
And when working on related sets of seeks, obviously use a seek list instead.
Then you'll need a text editor that can run sorts and line-counts easily. I can highly recommend Textpad (
Trim off and delete the header section of the export file that looks like this (hmm, I never noticed that the number of responses is actually given in the header...):
Brother Jerome * March 14 2006
29681.29674.34905 [UserID,MakerID,Responses]
StoryTeller: no
Compound: no
Memory: 3
Gossip: 0
AIScript: default "(name)" as "nickname";
...
Language Center
---------------
You will notice that all the keyphrases start at the left-hand margin, but responses are indented. Seeks are indented more heavily again, and so on.
Now select Sort in the tools menu (if you're using Textpad,) using only the first key, settings: "From: 1", "Descending", "Case insensitive", "in character code order".
Click OK
All the progressively more indenteded sections containing seeks, seeks of seeks, etc. will be sorted to the end of the file, and all the keyphrases to the beginning, in reverse alphabetical order. Scroll down to the end of the unindented section (keyphrases beginning with "("/numbers/"A",) and click on the last unindented line. The line number displayed in Textpad's status bar is the number of keyphrases, eg: 7320
Go to the last line of the next indented section, check the line number (eg: 42225) and subtract the number of keyphrases - this gives you the number of responses.
Repeat for each section you're interested in.
You can just highlight the section, and the line-count will appear in the status bar too.
Seeks aren't differentiated from their responses by an indent, but by begining with "+", so be careful when counting these.
I find a keyphrase list (which I would christen a "keynome"
) is an invaluable reference while working on your bot - it reduces redundant duplication, as well as giving a good general overview of where where you're at.And when working on related sets of seeks, obviously use a seek list instead.
I probably ought to mention that Textpad is fully-featured, time-unlimited shareware, available in a number of languages (including German,) with full spell-checking capabilities, and the registration is (I think) a very reasonable price for such a capable text editor (no, I'm not on commission, nor did I write it - I wish! )
)
You can also copy your export file into an Excel spreadsheet, that'll let you sort out your content as well by using filters and counts on each column.
» More new posts: Doghead's Cosmic Bar