Artificial Intelligence chatterbot and NLP (Natural Language Processing)

[Marcus]

What I'm trying to do is to use simple questions, that do not require very long or yes/no answers, yet can contain some keywords I can "hook" to. Your conversations do help to increase the 1-2 or even 1-2-3 dialogues, but some phrases either "loop" my thinking progress or grind it to halt.

Do you want me to stay with the more random "wide band" approach or go back over the territory I have aeady covered and see how she responds?

Its a pity I cant select the response with the mouse ... would save typing and give you the path you needed... Or can you mount and dismount sections of the DB before and after questions to limit the responses searched ?
Hmm I could thing of other areas the click selection could go ...smack ... bad Marcus 😋

[Like Ra]

Do you want me to stay with the more random "wide band" approach or go back over the territory I have aeady covered and see how she responds?

Both would be good. I need to go through my first approach (predetermined series of questions) again. Other wise it becomes "too routine". OTOH, some questions depend on others. That was my test of the tree-like question structure to get the most information out of the visitor to generate realistic and enjoyable self-bondage scenarios (not done yet). For example, if someone does not have any "professionally made toys", they should not be in the scenario. Also I tried to avoid repeating the same question, but the response must be properly understood and remembered.

would save typing and give you the path you needed...

I do not know what pass is chosen unless I enable the full debugging.There are 4 different priorities for found patterns (one of them does not work), and if there several finally found templates, one of them will be chosen randomly. And in between there can be multiple iterations of changing the user response, filling in variables, adding bot responses, etc. So, it's not easy and not quite predictable (what sometimes makes it more real).

Meanwhile you discovered a new Program O bug - three dots "..." is not recognized as an entries separator.

[Like Ra]

to generate realistic and enjoyable self-bondage scenarios

"Bondage generator" is simpler, because it should not be "real" in terms of available toys, clothes or physical capabilities 😉

[Marcus]

Hmmm Why don't you re-task that personality assessment code to do bondage preference assessment ? Or do you want to leverage that at a much deeper level later on in development ?

[Like Ra]

Why don't you re-task that personality assessment code to do bondage preference assessment ?

This is what I started with. The "usual" series of questions. But I'm not sure how to proceed. I can keep the predefined question tree (more boring) or make it random (though some questions depends on each other).

Or do you want to leverage that at a much deeper level later on in development ?

Yes, I'm going in this direction, but as I mentioned above, the "global structure" is not clear to me yet.

And I constantly stumble upon AIML limitations, ALICE AIML set limitations, program bugs, etc. For example, arrays do not exist. I cannot check if a variable is defined or not. I cannot use pattern matching for a variable. Pattern matching is severely limited (a part of a word cannot be used).

[Like Ra]

I changed the settings to save remember the visitors in the database. So, if you have the same IP, your conversation should start where you left it. All variables should be kept as they were set.

After seeing the latest Marcus' logs, I did not make any significant updates. I think I need to tweak the current logic by adding various patterns for the same templates. Actually, it's a nightmare.... Look at the "swimsuit" synonyms:

Code:

<category><pattern>* SWIMSUIT</pattern><template><srai>SWIMSUIT</srai></template></category>
<category><pattern>SWIMSUIT *</pattern><template><srai>SWIMSUIT</srai></template></category>
<category><pattern>* SWIM SUIT *</pattern><template><srai>SWIMSUIT</srai></template></category>
<category><pattern>_ SWIM SUIT</pattern><template><srai><star/>SWIMSUIT</srai></template></category>
<category><pattern>SWIM SUIT _</pattern><template><srai>SWIMSUIT<star/></srai></template></category>
<category><pattern>_ SWIM SUIT *</pattern><template><srai><star index="1"/>SWIMSUIT<star index="2"/></srai></template></category>
<category><pattern>* SWIMSUITS</pattern><template><srai><star/>SWIMSUIT</srai></template></category>
<category><pattern>SWIMSUITS *</pattern><template><srai>SWIMSUIT<star/></srai></template></category>
<category><pattern>* SWIMSUITS *</pattern><template><srai><star index="1"/>SWIMSUIT<star index="2"/></srai></template></category>

9 lines to cover for "swim suit", "swimsuits", "swimsuit" variations plus to check if there is something before or after... And this needs to be done for every "keyword", like lycra, spandex, pantyhose, latex, etc.

AIMLv2 will reduce the amount of the required synonyms by a factor of 3, but not many programs support v2.

Plurals and articles are killing me...

[Like Ra]

Readded the learning capability. It takes quite a while to learn a fact, but since it's learnt, it works fast again (thanks to added indexes, I think).

Just in case, can anybody ask "what is ab"? I wonder if it works "cross-session"

[Marcus]

I changed the settings to save remember the visitors in the database. So, if you have the same IP, your conversation should start where you left it. All variables should be kept as they were set.

It's not remembering ... seems like the IP address is now longer coming through like it used to ????

I Visited ALICE at google for a comparison .... yeah starting to appreciate the subtly of some of the "cheats".
BTW I think your occasional well chosen picture is streets ahead of animated avatars or fake Kirks.

[Like Ra]

Finding a suitable picture is easier than creating a suitable animated avatar. Though, should be possible especially in the anime style 😉 It can be, actually, an animated GIF with subtle movements. But in this case there is no much difference from the current approach. Same separate images, but animated.

[Like Ra]

Looks like the learning feature did something... Check this:

User: anyway, how are you
Miyuki: I enjoy completely closed hydrasuits, with a zip. I am functioning within normal parameters.
User: Completely closed?
Miyuki: By the way, I never liked bikini. I like one-piece swimsuits
User: I did not mention swimsuits
Miyuki: Why didn't you mention swimsuits?
User: Indeed, I do not know, you started talking about swimsuits
Miyuki: Who does know why you started talking about swimsuits?
User: No, you started talking about swimsuits
Miyuki: One-piece only. Preferably shiny.
User: And transparent
Miyuki: I also like transparent latex You are fully clothed, yet it looks like you are naked
User: Yes, this is very true. I also like transparent latex
Miyuki: Do any of you friends wear latex?
User: No, unfortunately not

Not bad, huh?