Home       Trees       Indices       Help   
Module model
[hide private]
[frames] | no frames]

Module model

source code


        Bachelor AI project 2009. Andreas van Cranenburgh 0440949.
        Two-word stage dialogue simulator using a corpus of exemplars.

        Interactive usage:
        $ python model.py

        Other uses:
                - Re-enact dialogue in 01.cha (should be a text file with one utterance
                  per line, with each line preceeded by *MOT or *CHI, as in the CHILDES
                  corpus):

                  $ python
#                 >>> import model
#                 >>> model.evaluate(open('01.cha').read())
                  ...

                - Talk to self for 25 utterances:

#                 >>> model.selftalk(25)
                  ...

                - Demonstrate some sample utterances that work (but not from corpus):

#                 >>> model.test()
                  ...
Functions [hide private]
 
arg(token) source code
 
conciliate(meaning, partialmeaning, subst=None, debug=<function dbg at 0x84bb64c>)
Test whether meaning can be made comptabible with partialmeaning, by looking for a family resemblance with partialmeanings, if found, perform a substitution if necessary.		 source code
 
dbg(*m)
Print debug output (also see nodbg).		 source code
 
dialogue(exemplars={}, constructions={}, lexicon={}, debug=<function dbg at 0x84bb64c>)
Co-routine for dialogue.		 source code
 
edit_dist(source, target)
Edit distance of two sequences.		 source code
 
evaluate(conv, exemplars={}, n=1)
When presented with a corpus fragment, feed parent utterances to model and generate model responses in place of child utterances.		 source code
 
express(meaning, exemplars, lexicon)
Express `meaning' by returning the most similar exemplar.		 source code
 
express2(meaning, exemplars, constructions, lexicon, debug=<function dbg at 0x84bb64c>)
Transform a meaning into a two word utterance using exemplars, constructions or the lexicon.		 source code
 
expressmulti(meaning, exemplars, constructions, lexicon)
Express `meaning' by returning a matching exemplar or construction		 source code
 
findmeaning(form, exemplars, lexicon, debug=<function dbg at 0x84bb64c>)
Given the words in a construction, find the most common meaning associated with it in the corpus of exemplars.		 source code
 
findtopic(discourse, debug=<function dbg at 0x84bb64c>)
Look for a recurring element in the last two utterances, if found, this becomes the new topic.		 source code
 
getexemplars()
Obtain corpus, either by importing from module `exemplars,' or by falling back to a small sample corpus.		 source code
 
inferconstructions(exemplars, lexicon, constructions={}, debug=<function dbg at 0x84bb64c>)
Build corpus of constructions from exemplars and lexicon.		 source code
 
inferlexicon(exemplars, verbose=False, lexicon={}, debug=<function dbg at 0x84bb64c>)
Infer lexicon from corpus of exemplars.		 source code
 
interpret(utterance, topic, exemplars, lexicon, debug=<function dbg at 0x84bb64c>)
Return semantic representation for linguistic utterance.		 source code
 
interpretwith(words, partialmeaning, exemplars, lexicon, debug=<function dbg at 0x84bb64c>)
Interpretation helper function called by interpret(), work out meaning of remaining words by stitching together the best matching fragments.		 source code
 
intersect(a, b) source code
 
lexiconcheck(utt, meaning, lex) source code
 
main()
Interactive textual user interface.		 source code
 
nodbg(*m) source code
 
pred(token) source code
 
reinforce(meaning, reaction, reactionutt, discourse, exemplars)
Strengthen connection between last two utterances by adding a random identifier and adding or updating the resulting exemplar.		 source code
 
response(meaning, exemplars, debug=<function dbg at 0x84bb64c>)
Transform a meaning into a response using adjacency pairs of speech acts.		 source code
 
revdict(d)
Reverse dictionary, ie.		 source code
 
selftalk(u=25, e={}, debug=<function dbg at 0x84bb64c>)
Talk to oneself, picking a random utterance from the exemplars when repetition is detected.		 source code
 
substr_iter(seq)
Return all substrings of a sequence, in descending order of length.		 source code
 
test()
Interpret some sample utterances.		 source code
 
tokens(m)
Turn meaning into a list of operator, predicates and arguments.		 source code
 
unifies(meaning, partialmeaning, debug=<function dbg at 0x84bb64c>)
succeed if everything in "partialmeaning" is compatible with "meaning", substituting along the way.		 source code
 
var(arg)
Test whether an argument string is variable:		 source code
 
varclauses(m) source code
Variables [hide private]
  __package__ = None
  lowercase = 'abcdefghijklmnopqrstuvwxyz'
  uppercase = 'ABCDEFGHIJKLMNOPQRSTUVWXYZ'
Function Details [hide private]

conciliate(meaning, partialmeaning, subst=None, debug=<function dbg at 0x84bb64c>)

source code 

Test whether meaning can be made comptabible with partialmeaning, by looking for a family resemblance with partialmeanings, if found, perform a substitution if necessary. Returns empty string on failure.

>>> conciliate('question: animal(bunny) do(X)', 'assertion: point(dog) animal(dog)')
substituted (bunny) for (dog)
'assertion: point(bunny) animal(bunny)'
>>> conciliate('assertion: do(hop) animal(bunny)', 'assertion: animal(bunny) do(X)')
instantiated (X) with (hop)
'assertion: animal(bunny) do(hop)'
>>> conciliate('assertion: do(hop) animal(bunny)', 'assertion: point(bunny)')
''

dbg(*m)

source code 

Print debug output (also see nodbg). Takes a variable number of arguments of any type.

>>> dbg('choices', range(3))
choices [0, 1, 2]

dialogue(exemplars={}, constructions={}, lexicon={}, debug=<function dbg at 0x84bb64c>)

source code 

Co-routine for dialogue. Use send(utterance) to receive a reply. Initial input: exemplars Further input: one utterance at a time Output: one reply at a time

edit_dist(source, target)

source code 

Edit distance of two sequences. Non-standard features are that all mutations have a cost of 1 (so as not to favor insertions or deletions over substitutions), as well as a decay based on index (mutations with higher indices weigh less).

>>> edit_dist('foo', 'bar')
2.4890507991136213

evaluate(conv, exemplars={}, n=1)

source code 

When presented with a corpus fragment, feed parent utterances to model and generate model responses in place of child utterances. n = number of iterations to perform Input: string with transcript of corpus Output: same transcript with replies from model instead of child

express2(meaning, exemplars, constructions, lexicon, debug=<function dbg at 0x84bb64c>)

source code 

Transform a meaning into a two word utterance using exemplars, constructions or the lexicon. Filter result according to lexical knowledge. Input: meaning representation Output: one or two word utterance

findtopic(discourse, debug=<function dbg at 0x84bb64c>)

source code 

Look for a recurring element in the last two utterances, if found, this becomes the new topic. Input: discourse (list of utterances and meanings) Output: clause, or None

inferconstructions(exemplars, lexicon, constructions={}, debug=<function dbg at 0x84bb64c>)

source code 

Build corpus of constructions from exemplars and lexicon. Input: exemplars & lexicon. Output: constructions, dictionary of pairings between substrings and clauses.

inferlexicon(exemplars, verbose=False, lexicon={}, debug=<function dbg at 0x84bb64c>)

source code 

Infer lexicon from corpus of exemplars. Input: exemplars. Output: lexicon, dictionary of word-clause pairings.

interpret(utterance, topic, exemplars, lexicon, debug=<function dbg at 0x84bb64c>)

source code 

Return semantic representation for linguistic utterance. This function is in charge of backtracking over initial exemplars and picking the best result, the rest is done by interpretwith() Input: utterance Output: meaning representation

reinforce(meaning, reaction, reactionutt, discourse, exemplars)

source code 

Strengthen connection between last two utterances by adding a random identifier and adding or updating the resulting exemplar. Input: two utterances with meanings Side-effect: updated or added exemplars

response(meaning, exemplars, debug=<function dbg at 0x84bb64c>)

source code 

Transform a meaning into a response using adjacency pairs of speech acts. Input: meaning representation Output: meaning representation

revdict(d)

source code 

Reverse dictionary, ie. swap values and keys; since a value may occur with multiple keys, return a list of all keys associated with a value.

>>> revdict({0: 1, 1: 2})
{1: [0], 2: [1]}

selftalk(u=25, e={}, debug=<function dbg at 0x84bb64c>)

source code 

Talk to oneself, picking a random utterance from the exemplars when repetition is detected. Input: number of utterances to generate, exemplars Output: transcript of conversation

substr_iter(seq)

source code 

Return all substrings of a sequence, in descending order of length.

>>> list(substr_iter('abc'))
['abc', 'ab', 'bc', 'a', 'b', 'c']

tokens(m)

source code 

Turn meaning into a list of operator, predicates and arguments.

>>> tokens('whquestion: do(X) animal(bunny)')
['whquestion:', 'do', '(X)', 'animal', '(bunny)']

unifies(meaning, partialmeaning, debug=<function dbg at 0x84bb64c>)

source code 

succeed if everything in "partialmeaning" is compatible with "meaning", substituting along the way. Similar to conciliate() but operates on the whole meaning instead of looking at individual clauses, and does not perform substitution, merely instantiation.

>>> unifies('assertion: do(hop) animal(bunny)', 'assertion: do(X) animal(bunny)')
substituted (hop) for (X)
'assertion: do(hop) animal(bunny)'
>>> unifies('assertion: do(hop) animal(bunny)', 'assertion: animal(bunny) do(X)')
''

var(arg)

source code 

Test whether an argument string is variable:

>>> var('(bar)')
False
>>> var('(X)')
True

   Home       Trees       Indices       Help   
Generated by Epydoc 3.0.1 on Sat Jun 27 18:12:56 2009 http://epydoc.sourceforge.net