The one being used in the video is chatbot-psi, more details can be found here: https://github.com/opencog/opencog/tree/master/opencog/nlp/chatbot-psi
I'm currently testing & fixing & adding new functions to "ghost", maybe not quite demo-able yet but hopefully it will soon be. It probably will replace chatbot-psi too. Remembering factual statements for question answering is possible, as well as calling PLN... On Mon, Aug 14, 2017 at 11:51 AM, Linas Vepstas <[email protected]> wrote: > p.s. That is Man Hin in the youtube video you attached. > > p.p.s. I was wrong, I can very quickly explain how the chatbot-old > question-answering system worked. It took IRC, parsed it with > link-grammar, and performed a pattern match on the grammatical parse of the > question. As long as there was some "fact" in the atomspace (i.e. bunch of > atoms organized in a grammatical fashion) then the pattern matcher could > find the "answer". Your demo "Tom eats apples. What does Tom eat?" are > grammatically trivial, and so are trivial to handle by grammatical pattern > matching. > > The new "ghost" system is kind-of like the old system, but on steroids. It > allows content author to author fairly complex answer-response systems. > Note, however, it is still an authoring system. > > I'm not sure if > > ghost has the ability to remember factual statements or not -- i.e. to do > question answering. Man Hin, can it do that? > > If it cannot do that, it should .. at the very lowest levels, its really > not hard. Its a good parlor trick. Its the handling of the sophisticated > stuff that becomes hard. > > I don't know if thee are plans for a common-sense database to be hooked > into ghost. I don't know if it will be possible to hook PLN into it. > > --linas > > > On Sun, Aug 13, 2017 at 8:20 PM, Linas Vepstas <[email protected]> > wrote: > >> That interface is a mostly broken experimental scaffolding where ... >> somethings work and most things don't. I don't think its maintained; I >> was probably the last person to mess with it, maybe 6 months ago. I still >> have plans to do stuff with it, but that's just me and my personal plans. >> >> Meanwhile, there are several other chatbot interfaces maintained by >> various people. Not sure where they are, what the status is, or who is >> doing what. Man Han has been working on "ghost" but don't know if that can >> be demoed or not. >> >> This dialog segment: Tom eats apples. What does Tom eat? is a typical >> question-answering system. We used to have one of those, about 8 years ago, >> and it mostly worked... it probably could have answered your questions. It >> was abandoned because of multiple reasons, some technical, some political. >> You can look at that old code here: https://github.com/opencog/ope >> ncog/tree/master/opencog/nlp/chatbot-old/chatbot >> >> That old code is not worth reviving, however, the lessons learned there >> were very very useful (to me) and helped clarify various core issues >> surrounding knowleege, common sense and learning. I cannot transmit those >> lessons to you quickly and easily, (but I can transmit them, with effort). >> >> Short answer: its not hard to create question-answering systems. But >> that's not where the action is. On the other hand, we (someone) really >> should get the basic question-answering scaffolding working again, on >> opencog. Its a shame that it was allowed to bit-rot. >> >> --linas >> >> On Fri, Aug 11, 2017 at 4:56 AM, Pavel Savyuk <[email protected]> >> wrote: >> >>> Hi All, >>> I'm trying to use the chatbot (in nlp/chatbot), using the sample dialog >>> provided on the GitHub page for it, and I seem to not be receiving replies. >>> Specifically: >>> >>> - Whenever I make a Declarative SpeechAct, I get a confirmation (I >>> assume this is normal behavior) >>> - Whenever I make an Interrogative SpeechAct, I get a list of >>> everything I said as an array (sometimes the reply is instant, sometimes >>> it's delayed by several seconds with CPU usage spiking, so I assume >>> something is going on...) >>> - Whenever I make a Truth Query, I get "I can't process Truth Query >>> for now" >>> >>> eva-chatbot behavior is the same, but it correctly responds to the robot >>> manipulation commands (e.g., look left) >>> >>> What am I doing wrong? I want to get the possibility of a dialog such as >>> in the "Simple OpenCog inference" video @ https://www.youtube.com/watc >>> h?v=LduD7Et_cOs >>> My commands are below. Thanks for any help! >>> >>> Terminal #1 >>> pavel@opencog2:~/relex$ ./opencog-server.sh >>> [...] >>> Info: Waiting for socket connection >>> >>> Terminal #2 >>> pavel@opencog2:~/opencog/opencog/nlp/chatbot$ guile -l run-chatbot.scm >>> [...] >>> scheme@(guile-user)> >>> >>> Terminal #3 >>> pavel@opencog2:~$ rlwrap telnet localhost 17004 >>> Trying ::1... >>> Trying 127.0.0.1... >>> Connected to localhost. >>> Escape character is '^]'. >>> (process-query "Pavel" "Tom eats apples.") >>> Entering scheme shell; use ^D or a single . on a line by itself to exit. >>> guile> Hello Pavel, you said: "Tom eats apples." >>> You made a Declarative SpeechAct >>> guile> (process-query "Pavel" "What does Tom eat?") >>> Hello Pavel, you said: "What does Tom eat?" >>> You made an Interrogative SpeechAct >>> (Tom eats apples . what does Tom eat ?) >>> guile> (process-query "Pavel" "What does Tom drink?") >>> Hello Pavel, you said: "What does Tom drink?" >>> You made an Interrogative SpeechAct >>> (Tom eats apples . what does Tom eat ? what does Tom drink ?) >>> guile> (process-query "Pavel" "What does Tom eat?") >>> Hello Pavel, you said: "What does Tom eat?" >>> You made an Interrogative SpeechAct >>> (Tom eats apples . what does Tom eat ? what does Tom drink ?) >>> guile> >>> >>> >>> -- >>> You received this message because you are subscribed to the Google >>> Groups "opencog" group. >>> To unsubscribe from this group and stop receiving emails from it, send >>> an email to [email protected]. >>> To post to this group, send email to [email protected]. >>> Visit this group at https://groups.google.com/group/opencog. >>> To view this discussion on the web visit https://groups.google.com/d/ms >>> gid/opencog/5c2d5140-ba98-40d8-ae30-5f0d2a63a47c%40googlegroups.com >>> <https://groups.google.com/d/msgid/opencog/5c2d5140-ba98-40d8-ae30-5f0d2a63a47c%40googlegroups.com?utm_medium=email&utm_source=footer> >>> . >>> For more options, visit https://groups.google.com/d/optout. >>> >> >> > -- > You received this message because you are subscribed to the Google Groups > "opencog" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To post to this group, send email to [email protected]. > Visit this group at https://groups.google.com/group/opencog. > To view this discussion on the web visit https://groups.google.com/d/ms > gid/opencog/CAHrUA34dSKD%3DEEevb5qKj-bak0ju1XZfkN5frUfqq9bfG > %3DG9EQ%40mail.gmail.com > <https://groups.google.com/d/msgid/opencog/CAHrUA34dSKD%3DEEevb5qKj-bak0ju1XZfkN5frUfqq9bfG%3DG9EQ%40mail.gmail.com?utm_medium=email&utm_source=footer> > . > > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "opencog" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at https://groups.google.com/group/opencog. To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CAMfi0BJTeK72iTw8_Y2bgoED21hGmWf-j%3Dj8iXNrwME%2BeXwfJg%40mail.gmail.com. For more options, visit https://groups.google.com/d/optout.
