Posts: 25,780
Thanks received: 12780
Thanks given: 6890
Joined: Jul 2006
(20 Feb 2015, 23:16 )Marcus Wrote: Q. Does she actually gossip between sessions/chats ? That was unexpected and some how disturbingly human....
AFAIK, AIML allows such things, but it depends on the implementation. I do not think it's properly implemented in Program O (if at all).
(20 Feb 2015, 23:16 )Marcus Wrote: Q. Is she remembering sessions/people by IP address yet ?
I set it up as "session only", but looks like she remembers everything 😁
(20 Feb 2015, 23:16 )Marcus Wrote: She is actually getting better... Its still a conscious effort to hold her on conversation... but she is miles ahead of where she started.
I added several thousands of lines and changed hundreds of default responses. So it should be a bit different.
(20 Feb 2015, 23:16 )Marcus Wrote: I'm not sure the joke thing is so good...
I removed one of the "humor" files, what fixed (I hope!) that "bandaide" problem 😉
Posts: 25,780
Thanks received: 12780
Thanks given: 6890
Joined: Jul 2006
(20 Feb 2015, 23:16 )Marcus Wrote: Mind you I did have several WOW moment responses in that session.
YESSSSSSSSSSSSSS!!!!!!!!!!!! 😁
Posts: 122
Thanks received: 11
Thanks given: 60
Joined: Mar 2010
21 Feb 2015, 11:22
(This post was last modified: 21 Feb 2015, 11:40 by Marcus.)
(20 Feb 2015, 23:49 )Like Ra Wrote: (20 Feb 2015, 23:16 )Marcus Wrote: Mind you I did have several WOW moment responses in that session.
YESSSSSSSSSSSSSS!!!!!!!!!!!! 😁
She is kinda temperamental ... Is this too much data or just interactions ?
There is good depth... but the kinkiness needs to be more subtly systemic...
I think the context is not matching the response ???
If I choose my words right she behaves well...
She does need a swimsuit picture... this is LikeRa after all....
If possible she should make you ask a few times before she shows you ? (I assume she is a tease?)
That image timing is spot on ... that works great... feels right. Well done !!
She customised my bondage based on the conversation .... very well done !!! Gave me a mental hard on 😋
Just a thought ... she shouldn't tie people up till she knows 3 or 4 things about them...
Posts: 25,780
Thanks received: 12780
Thanks given: 6890
Joined: Jul 2006
21 Feb 2015, 13:10
(This post was last modified: 21 Feb 2015, 13:11 by Like Ra.)
Yes, these "3-4" or "few times" can be programmed in AIML using many variables and conditions, but it would make the code overly complex. Another possibility is to use JavaScript, however I haven't tried to exchange the data between AIML and JS. Currently I use JS "passively" to show the pictures. But sooner or later I will come to it.
My current priority is to fill as much "fetish gaps" and add multiple responses as possible. The first steps is "wide", the next is "a bit deeper", the 3d one is more "real" interaction when the bot follows the subject.
What I noticed is that the ALICE files were not properly looked at by linguists. I'm not a linguist, but I do see obvious bugs. For example, that first thing that should be eliminated in English is articles. All these "A", "AN", "THE" add enormous redundancy. Fortunately, there are no cases (unlike, for example, in German) or male/female forms.
Another interesting idea is to randomly choose between predefined synonyms, when possible.
Unfortunately, the AIML standard (even v2) does not offer much flexibility in patter matching, but creating external pre-parsers is a completely different project 😉
Posts: 25,780
Thanks received: 12780
Thanks given: 6890
Joined: Jul 2006
BTW, I've just checked the last transcripts, and it looks like either the database, or the caching, or the Program O itself has an annoying bug - it picks up one line and uses it as a sort of "default response" for short messages like "Yes" or "No". Before it was "bandaide", now it's "I'm a kinky schoolgirl". These strings are completely irrelevant. Probably I need to increase the debugging level...
Oh, what am I busy with ... ? ...
Posts: 25,780
Thanks received: 12780
Thanks given: 6890
Joined: Jul 2006
Ha! I found it! Finally! Fixed! Oh... Why didn't I notice that before.... 😁
The next thing to fix is the "Bound" / "Both" answers. They do not reach the destination.
Step by step ... 😉
Posts: 25,780
Thanks received: 12780
Thanks given: 6890
Joined: Jul 2006
Did not add too many new features but I did fix many bugs today 😉 Also, tried to play with JavaScript, but it turned out that it cannot be used to effectively parse the text, because the result cannot be returned back for further processing. But I tested the possibility to add external UNIX programs and that worked! 😉
Posts: 25,780
Thanks received: 12780
Thanks given: 6890
Joined: Jul 2006
Program O does not like punctuations. For example, it cannot parse words with dashes. E.g. "selfbondage" works, "self bondage" works, "self-bondage" does not.
Posts: 25,780
Thanks received: 12780
Thanks given: 6890
Joined: Jul 2006
Solved this by using the built into Program O customizable spell-corrector. Now it works faster and more precise 😉
Posts: 122
Thanks received: 11
Thanks given: 60
Joined: Mar 2010
Thanks Ra ..... Love her swimsuit 😋
She is becoming quite a challenge to explore....
She is complex but gives enough hints... like a real person flirting.... 💓
O well I can dream and play ....
Was quite amazed at some of the responses today.
The change of topic and circling back with a question much later is making her seem more real.
I think the closer she gets the more disappointed I get when she breaks the illusion that she is human.
Q. If your her Father who is the Mother ?
