|
|
Senior member
Total posts: 136
Joined: Jun 20, 2014
|
Does CS currently support any way to add patterns and responses without hardcoding them? Maybe by learning from what is said in a conversation?
Something like:
user: Do you know what * is
CS : no do you?
user: * is X
then later when asked the same question can respond with an answer instead of a question
One way i thought to do this was by checking for a fact regarding *, if it knew it respond with that otherwise ask and store the fact if the pattern matched.
Still learning facts and macros so not sure about this. Just wondering what options CS has for learning.
|
|
|
|
|
Posted: Mar 31, 2015 |
[ # 1 ]
|
|
Moderator
Total posts: 2372
Joined: Jan 12, 2010
|
Nominally, learning would be done not as a pattern and response, but as a fact saved and used by patterns.
u: ( do you know what _*1 is ) query(direct_sv _0 information ?)
_0 is @0object .
u: ( do you know what _*1 is) $what = ‘_0 No, Do you?
a: (_*1 ‘_0=$what is _* ) createfact ( $what information ‘_1)
|
|
|
|
|
Posted: Mar 31, 2015 |
[ # 2 ]
|
|
Senior member
Total posts: 136
Joined: Jun 20, 2014
|
‘Course that would mean a huge fact database. But depending on the type of questions(why where what etc.) and answers you could start building a database similar to the way children ask questions and learn. Can a fact item be more than one word? like (“Michael Jordan” info “basketball player”)
|
|
|
|
|
Posted: Mar 31, 2015 |
[ # 3 ]
|
|
Moderator
Total posts: 2372
Joined: Jan 12, 2010
|
if you get to a big database, then switch to storing data in postgres.
A fact field can be an entire sentence.
u: (_*) createfact (‘_0 input sentence)
|
|
|
|
|
Posted: Mar 31, 2015 |
[ # 4 ]
|
|
Senior member
Total posts: 136
Joined: Jun 20, 2014
|
Thanks Bruce. Ive never used postgres but I am familiar with sql.
Can I store a null value as a fact field? And is there any way to deal with transcripts of chats using chatscript?
|
|
|
|
|
Posted: Mar 31, 2015 |
[ # 5 ]
|
|
Moderator
Total posts: 2372
Joined: Jan 12, 2010
|
|
|
|
|
|
Posted: Mar 31, 2015 |
[ # 6 ]
|
|
Senior member
Total posts: 136
Joined: Jun 20, 2014
|
To both questions? I may have added the second after editing.
|
|
|
|
|
Posted: Mar 31, 2015 |
[ # 7 ]
|
|
Moderator
Total posts: 2372
Joined: Jan 12, 2010
|
Ah. dealing with transcripts of chats…. what did you have in mind?
:trim can be used to improve human readability.
:document can process text documents.
|
|
|
|
|
Posted: Mar 31, 2015 |
[ # 8 ]
|
|
Senior member
Total posts: 136
Joined: Jun 20, 2014
|
Just thinking of ideas since CS seems so versatile. I have lots of time to kill but find it hard to commit to something.
Also, is there a way to define patterns by by their pos type?
Like (a {ADJ}NOUN VERB because *)
That might be useful for learning the answer to why questions
|
|
|
|
|
Posted: Mar 31, 2015 |
[ # 9 ]
|
|
Moderator
Total posts: 2372
Joined: Jan 12, 2010
|
estorica folder has posparser documentation….
u: ( ~determiner {~adjective} ~noun ~verb )
u: (~noun_singular ~verb_past )
|
|
|
|
|
Posted: Mar 31, 2015 |
[ # 10 ]
|
|
Senior member
Total posts: 136
Joined: Jun 20, 2014
|
Btw, Pandorabots is the only site I know that hosts chatbots and oddly enough they seem to have no support for CS. Are you aware of any sites that do?
|
|
|
|
|
Posted: Mar 31, 2015 |
[ # 11 ]
|
|
Moderator
Total posts: 2372
Joined: Jan 12, 2010
|
No. There is no general service for hosting multiple CS bots.
There are other sites than Pandora (AIML), PersonalityForge (their own language), and various others of similar ilk to aiml.
|
|
|
|
|
Posted: Mar 31, 2015 |
[ # 12 ]
|
|
Senior member
Total posts: 136
Joined: Jun 20, 2014
|
Bah. Once you’ve experienced CS you can’t go back to AIML.
|
|
|
|