a really simple chat bot
Chortle is still very new to the human world and will take a while to get used to human conversation. Chortle will (often) listen to what a human says in conversation, but will not always know the best way to respond. For some words or phrases, however, Chortle has been taught common, contextual responses by a teacher in an attempt to help Chortle along the path of becoming a better conversationalist.
In conversation (chat mode), Chortle:
- knows when a human is making a statement
- knows when a human is asking a question
- will respond with learned phrases
- will respond with new response guesses
- has some questions to ask human
- will remember human answers to questions (relational data)
- can recall human answers to questions (relational data)
- has its own favorites (relational data)
- can respond in kind if human relational data matches Chortle relational data
In learning (teacher mode), Chortle:
- waits for teacher to say a phrase
- listens to phrase and responds with a related phrase guess (if available)
- will take a guess if no previously-learned response is found
- waits for teacher response (yes: good response ... no: bad response)
- increases weight (usefulness) of response phrase, depending on teacher's response decision
- decreases weight (usefulness) of response phrase, depending on teacher's response decision
- will forget responses that are no longer useful (weight pruning)
- "What is your name?"
- "What are your hobbies?"
- ...and whatever else comes to mind! (check out the chortle-conversations archive for past gems...)
__ __ __
____ | |__ _____ ____ | |_| | __?__
| __|| | _ | _|| _| || -__|
|____||__|__|_____|__| |____|__||_____|
1) chat mode
2) teacher mode
3) quit
1
bot > Lovely day, isn't it?
bot > What is your name?
human > Jonathan
bot > Nice to meet you!
human > what is your name?
bot > chortle
human > nice to meet you
human > what are your hobbies?
bot > chatting
human > oh really
human > yes
bot > Oh, how lovely!
human > haha
bot > Do you like fruit?
human > yes
bot > Ah, I see.
human > that's cool
human > do you like food?
bot > What is your favorite food?
human > i like pizza
bot > What do you mean?
human > oh, a food that is bad for you
bot > Ah, I see.
human > yeah
bot > Be seeing you!
human > oh, okay
bot > Ah, I see.
human > bye
bot > Well, I need to be going!
Want to see what I learned about you?
yes
learned relational data:
your name
is
Jonathan
you
like
fruit
your favorite food
is
pizza
push enter to continue
bot > What is your name?
human > my name is Jonathan
bot > I see
my name is Jonathan
found final verb
generated key/value lists joined individually
PRP,UNKNOWN,VBZ
Jonathan,bean
UNKNOWN,UNKNOWN
>>> result: Jonathan
bot > What is your favorite food?
human > green jelly
bot > I see
green jelly
>>> result: green jelly
bot > What is your favorite color?
human > red ruby
bot > I see
red ruby
>>> result: red ruby
bot > Do you like red ruby green jelly?
human > yes
bot > I see
yes
>>> result: yes
learned information:
your name - Jonathan
your favorite color - red ruby
your favorite food - green jelly
you like red ruby green jelly - yes
The final verb check is used to test if Chortle can find the intended target verb in a given phrase (where there are two actual verbs provided, and the intended target verb is the final verb in given phrase.) Once Chortle decides which verb to use, Chortle is able to parse out the intended subject, verb, and object of the given phrase.
bot > what is the way to go? (multiple verb test)
human > my way to go is over there
bot > I see
responsePOS: PRP,UNKNOWN,UNKNOWN,VBZ,VBZ,UNKNOWN,UNKNOWN
my way to go is over there
> found match!
found final verb: VBZ
generated key/value lists joined individually
PRP,UNKNOWN,UNKNOWN,VBZ,VBZ
over,there
UNKNOWN,UNKNOWN
>>> result: over there
learned information:
way to go - over there
The DT_X_VB check is used to test if Chortle can find the correct object when a determiner is being used near the beginning of a phrase. For example, the initial question given by bot could be in a pattern such as: O-V-S (object-verb-subject). So the human response may be given in a similar O-V-S pattern... and if so, Chortle needs to be able to understand what the human is using for the subject and object of the overall phrase.
bot > What is the way to go? (multiple verb test)
human > that way is the way
bot > I see
responsePOS: DT,UNKNOWN,VBZ,DT,UNKNOWN
> found DT_X_VB!
that way is the way
> found match!
right to left verb direction!
found root verb: VBZ
generated key/value lists joined individually
generated key list: the,way
generated value list: that,way
generated value pattern list: DT,UNKNOWN
>>> result: that way
learned information:
way to go - that way
teacher > hello
bot > See ya!
teacher > 1:yes, 2:no > 2
teacher > hello
bot > Lovely day, isn't it?
teacher > 1:yes, 2:no > 1
teacher > goodbye
bot > Oh, hey there!
teacher > 1:yes, 2:no > 2
teacher > goodbye
bot > Be seeing you!
teacher > 1:yes, 2:no > 1
teacher > goodbye
bot > Be seeing you!
teacher > 1:yes, 2:no > 1
learned responses:
*
[What do you mean?, 0.5]
[Hmm?, 0.5]
hello
[Oh, hey there!, 0.5]
[Lovely day, isn't it?, 0.6]
[Be seeing you!, 0.5]
[See ya!, 0.4]
[Nice to meet you!, 0.5]
goodbye
[Oh, hey there!, 0.4]
[Lovely day, isn't it?, 0.5]
[Be seeing you!, 0.7]
[See ya!, 0.5]
[Nice to meet you!, 0.5]
i like
[Good for you!, 0.5]
[I'm sorry to hear that., 0.5]
[Oh, how lovely!, 0.5]
[Sounds great., 0.5]
[Yes, I suppose you must be., 0.5]
[Wow! That sounds exciting., 0.5]
my name is
[Oh, hey there!, 0.5]
[Lovely day, isn't it?, 0.5]
[Be seeing you!, 0.5]
[See ya!, 0.5]
[Nice to meet you!, 0.5]
thanks
[You are welcome!, 0.5]
[Don't mention it., 0.5]
[No problem., 0.5]
[It's okay, 0.5]
teacher > hello there
> finding response...
> trying shorter key: hello there
> found a response: False
> picking from grab bag...
> checking weight response...0.5 / What do you mean?
> found an okay weight response
> checking weight response...0.5 / Hmm?
> found an okay weight response
> finding response...
> trying shorter key: hello
> found this shorter key in learned responses: hello
> checking weight response...0.5 / 0.5
> checking an 'okay' response
> found an okay weight response
> checking weight response...0.5 / 0.5
> checking an 'okay' response
> found an okay weight response
> checking weight response...0.5 / 0.5
> checking an 'okay' response
> found an okay weight response
> checking weight response...0.5 / 0.5
> checking an 'okay' response
> found an okay weight response
> checking weight response...0.5 / 0.5
> checking an 'okay' response
> found an okay weight response
> botResponse: Nice to meet you!
> found a response: False
> picking from grab bag...
> checking weight response...0.5 / What do you mean?
> found an okay weight response
> checking weight response...0.5 / Hmm?
> found an okay weight response
> finding response...
> trying shorter key:
> found a response: False
> picking from grab bag...
> checking weight response...0.5 / What do you mean?
> found an okay weight response
> checking weight response...0.5 / Hmm?
> found an okay weight response
bot > Hmm?
teacher > 1:yes, 2:no > 2
> adding new bot response key: Hmm?
teacher > hello there
> found key: hello there
> checking weight response...0.4 / Hmm?
> found response: Nice to meet you!
bot > Nice to meet you!
teacher > 1:yes, 2:no > 1
> adding response to learned key ...
teacher > hello there
> found key: hello there
> checking weight response...0.4 / Hmm?
> found response: Wow! That sounds exciting.
> checking weight response...0.6 / Nice to meet you!
> found a good weight response
bot > Nice to meet you!
teacher > 1:yes, 2:no > 2
> bot already knows this response... but let's update info
teacher > hello there
> found key: hello there
> checking weight response...0.5 / Nice to meet you!
> found an okay weight response
> checking weight response...0.4 / Hmm?
> found response: Nice to meet you!
bot > Nice to meet you!
teacher > 1:yes, 2:no > 2
> bot already knows this response... but let's update info
teacher > hello there
> found key: hello there
> checking weight response...0.4 / Hmm?
> found response: Nice to meet you!
> checking weight response...0.4 / Nice to meet you!
> found response: Oh, hey there!
bot > Oh, hey there!
teacher > 1:yes, 2:no > 1
> adding response to learned key ...
learned responses:
*
[What do you mean?, 0.5]
[Hmm?, 0.5]
hello
[Oh, hey there!, 0.5]
[Lovely day, isn't it?, 0.5]
[Be seeing you!, 0.5]
[See ya!, 0.5]
[Nice to meet you!, 0.5]
goodbye
[Oh, hey there!, 0.5]
[Lovely day, isn't it?, 0.5]
[Be seeing you!, 0.5]
[See ya!, 0.5]
[Nice to meet you!, 0.5]
i like
[Good for you!, 0.5]
[I'm sorry to hear that., 0.5]
[Oh, how lovely!, 0.5]
[Sounds great., 0.5]
[Yes, I suppose you must be., 0.5]
[Wow! That sounds exciting., 0.5]
my name is
[Oh, hey there!, 0.5]
[Lovely day, isn't it?, 0.5]
[Be seeing you!, 0.5]
[See ya!, 0.5]
[Nice to meet you!, 0.5]
thanks
[You are welcome!, 0.5]
[Don't mention it., 0.5]
[No problem., 0.5]
[It's okay, 0.5]
hello there
[Hmm?, 0.4]
[Nice to meet you!, 0.4]
[Oh, hey there!, 0.6]
- some conversation response data comes from: http://en.wikibooks.org/wiki/English_in_Use/Conversation_Pieces
- parts of speech (POS) tags adhere to Illinois POS: http://cogcomp.cs.illinois.edu/demo/pos/?id=4
- POS key is located in pos-key.txt for general reference