Artificial Intelligence chatterbot and NLP (Natural Language Processing)
302 Replies, 91769 Views
(13 Feb 2015, 14:34 )Like Ra Wrote:If the pictures where indexed by keyword and those words used as the lookup on the trigger? Ie the response text it's self picks/indexes/lockup the picture so very little work in programming. The approach has limitations but would work as a proof of concept with reasonable depth. Obviously a random element to resolve the 1 to many lookups.(13 Feb 2015, 03:11 )Marcus Wrote: The picture responses need to be a bit like a snap chatCan be a good idea, but in this case I will need to insert a javascript line with a separate picture almost in every pattern. Way too much work. Anyhow time for my daily chat with Alice....
(14 Feb 2015, 05:04 )Marcus Wrote: If the pictures where indexed by keyword and those words used as the lookup on the trigger? IHa! This is what I mentioned earlier in this thread. Tagging and indexing are extremely important and very difficult to properly implement and maintain. I tried to implement tagging here in the forum, but it did not work properly. Google does this automatically using image recognition and neural networks. To do this manually - too much work for one person. Even blog tagging is difficult 😉
But I have a couple of ideas 😉
OK, one of the ideas costs money:
Quote:Custom Search Engine (free)
This dead loops has been fixed:
Quote:Miyuki: I do not completely understand if it was "yes" or "no"? Another attempt Thanks for testing! But I still do not understand, while the program keeps ignoring several statements...
Got two bugs confirmed (not fixed yet!) by Program O developers. And logged another one:
(This post was last modified: 19 Feb 2015, 23:07 by Like Ra.)
User: Hello my name is Like Ra Miyuki: Nice to meet you, Like Ra User: Hello, my name is Like Ra Miyuki: There's no other girl like me. The comma should be ignored, but it prevents the correct pattern to match. I stumbled upon more unexpected behaviours, but I don't know how to properly formulate them yet. I managed to create some complex self-generating trees, but the actual user interaction still fails to work properly.
Marcus, thanks for testing. I know about this issue:
marcus: yes Miyuki: bandaide. marcus: 😟 But I have NO IDEA where it comes from... I've been trying to fix that for days, actually. Anyway, how do you like some new features you managed to discover? 😉
And I must say that I love your dialogues! 😁
(20 Feb 2015, 22:49 )Like Ra Wrote: Marcus, thanks for testing. I know about this issue: She is actually getting better... Its still a conscious effort to hold her on conversation... but she is miles ahead of where she started. Nice that she has started to remember things now ... that's a nice touch... it doesn't have to be much to make a big difference. She is also starting to hold topics too I see... I'm not sure the joke thing is so good... well it needs the right jokes, she prolly needs to joke (poke fun at) at herself ? Q. Does she actually gossip between sessions/chats ? That was unexpected and some how disturbingly human.... She asked permission to tell the jokes and use the gossip... a real person would just do it...If she had just "done it" that would have been a real WOW moment. Mind you I did have several WOW moment responses in that session. The bondage pics (while great) really only want to stay up for one or two cycles then back to her avatar. As you pointed out its her I'm engaging not a photo album. Q. Is she remembering sessions/people by IP address yet ? Once she is sentient you can download her to a chassis and let her loose on the world 😉 Cant wait to "comb her wires" again.... |
