def display_meaning(word):
meaning = get_meaning(word)
window['output'].print("WORD: " + word)
if meaning:
window['output'].print("MEANING: ", meaning + ".")
else:
display_error("Word is not found in corpus.")
def get_sentence_posteriors(sentence, iterations=1, extra_meaning=None):
meaning_probs = {}
# parse sentence with charniak and apply surgeries
print 'parsing ...'
modparse = get_modparse(sentence)
t = ParentedTree.parse(modparse)
print '\n%s\n' % t.pprint()
num_ancestors = count_lmk_phrases(t) - 1
for _ in xrange(iterations):
(lmk, _, _), (rel, _, _) = get_meaning(num_ancestors=num_ancestors)
meaning = m2s(lmk,rel)
if meaning not in meaning_probs:
ps = get_tree_probs(t, lmk, rel)[0]
# print "Tree probs: ", zip(ps,rls)
meaning_probs[meaning] = np.prod(ps)
print '.'
if extra_meaning:
meaning = m2s(*extra_meaning)
if meaning not in meaning_probs:
ps = get_tree_probs(t, lmk, rel)[0]
# print "Tree prob: ", zip(ps,rls)
meaning_probs[meaning] = np.prod(ps)
print '.'
summ = sum(meaning_probs.values())
for key in meaning_probs:
meaning_probs[key] /= summ
return meaning_probs.items()
def display_meaning(word):
meaning=get_meaning(word)
window["output"].print("WORD:",word)
if meaning:
window["output"].print("MEANING:",meaning)
else:
display_error("word in not found in corpus")
def display_meaning(word):
meaning = get_meaning(word)
window['output'].print("Word: " + word)
if meaning:
window['output'].print("Meaning: ", meaning)
else:
display_error("Meaning Not Found")
def generate_sentence(loc, consistent, scene=None, speaker=None):
utils.scene = utils.ModelScene(scene, speaker)
(lmk, lmk_prob, lmk_ent), (rel, rel_prob, rel_ent) = get_meaning(loc=loc)
meaning1 = m2s(lmk, rel)
logger(meaning1)
while True:
rel_exp_chain, rele_prob_chain, rele_ent_chain, rel_terminals, rel_landmarks = get_expansion(
'RELATION', rel=rel)
lmk_exp_chain, lmke_prob_chain, lmke_ent_chain, lmk_terminals, lmk_landmarks = get_expansion(
'LANDMARK-PHRASE', lmk=lmk)
rel_words, relw_prob, relw_ent, rel_a = get_words(
rel_terminals, landmarks=rel_landmarks, rel=rel)
lmk_words, lmkw_prob, lmkw_ent, lmk_a = get_words(
lmk_terminals,
landmarks=lmk_landmarks,
prevword=(rel_words[-1] if rel_words else None))
sentence = ' '.join(rel_words + lmk_words)
logger('rel_exp_chain: %s' % rel_exp_chain)
logger('lmk_exp_chain: %s' % lmk_exp_chain)
meaning = Meaning(
(lmk, lmk_prob, lmk_ent, rel, rel_prob, rel_ent, rel_exp_chain,
rele_prob_chain, rele_ent_chain, rel_terminals, rel_landmarks,
lmk_exp_chain, lmke_prob_chain, lmke_ent_chain, lmk_terminals,
lmk_landmarks, rel_words, relw_prob, relw_ent, lmk_words,
lmkw_prob, lmkw_ent))
meaning.rel_a = rel_a
meaning.lmk_a = lmk_a
if consistent:
# get the most likely meaning for the generated sentence
try:
posteriors = get_sentence_posteriors(sentence,
iterations=10,
extra_meaning=(lmk, rel))
except:
print 'try again ...'
continue
meaning2 = max(posteriors, key=itemgetter(1))[0]
# is the original meaning the best one?
if meaning1 != meaning2:
print
print 'sentence:', sentence
print 'original:', meaning1
print 'interpreted:', meaning2
print 'try again ...'
print
continue
for m, p in sorted(posteriors, key=itemgetter(1)):
print m, p
return meaning, sentence
def generate_sentence(loc, consistent, scene=None, speaker=None, usebest=False, golden=False, meaning=None, printing=True):
utils.scene = utils.ModelScene(scene, speaker)
if meaning is None:
(lmk, lmk_prob, lmk_ent), (rel, rel_prob, rel_ent) = get_meaning(loc=loc, usebest=usebest)
else:
lmk, rel = meaning
lmk_prob = lmk_ent = rel_prob = rel_ent = None
meaning1 = m2s(lmk, rel)
logger( meaning1 )
while True:
rel_exp_chain, rele_prob_chain, rele_ent_chain, rel_terminals, rel_landmarks = get_expansion('RELATION', rel=rel, usebest=usebest, golden=golden, printing=printing)
lmk_exp_chain, lmke_prob_chain, lmke_ent_chain, lmk_terminals, lmk_landmarks = get_expansion('LANDMARK-PHRASE', lmk=lmk, usebest=usebest, golden=golden, printing=printing)
rel_words, relw_prob, relw_ent, rel_a = get_words(rel_terminals, landmarks=rel_landmarks, rel=rel, usebest=usebest, golden=golden, printing=printing)
lmk_words, lmkw_prob, lmkw_ent, lmk_a = get_words(lmk_terminals, landmarks=lmk_landmarks, prevword=(rel_words[-1] if rel_words else None), usebest=usebest, golden=golden, printing=printing)
sentence = ' '.join(rel_words + lmk_words)
if printing: logger( 'rel_exp_chain: %s' % rel_exp_chain )
if printing: logger( 'lmk_exp_chain: %s' % lmk_exp_chain )
meaning = Meaning((lmk, lmk_prob, lmk_ent,
rel, rel_prob, rel_ent,
rel_exp_chain, rele_prob_chain, rele_ent_chain, rel_terminals, rel_landmarks,
lmk_exp_chain, lmke_prob_chain, lmke_ent_chain, lmk_terminals, lmk_landmarks,
rel_words, relw_prob, relw_ent,
lmk_words, lmkw_prob, lmkw_ent))
meaning.rel_a = rel_a
meaning.lmk_a = lmk_a
if consistent:
# get the most likely meaning for the generated sentence
try:
posteriors = get_sentence_posteriors(sentence, iterations=10, extra_meaning=(lmk,rel))
except:
print 'try again ...'
continue
meaning2 = max(posteriors, key=itemgetter(1))[0]
# is the original meaning the best one?
if meaning1 != meaning2:
print
print 'sentence:', sentence
print 'original:', meaning1
print 'interpreted:', meaning2
print 'try again ...'
print
continue
for m,p in sorted(posteriors, key=itemgetter(1)):
print m, p
return meaning, sentence
def generate_sentence(loc, consistent):
while True:
lmk, rel = get_meaning(loc=loc)
print m2s(lmk, rel)
rel_exp, rel_prob, rel_ent = get_expansion('RELATION', rel=rel)
lmk_exp, lmk_prob, lmk_ent = get_expansion('LANDMARK-PHRASE', lmk=lmk)
rel_words, relw_prob, relw_ent = get_words(rel_exp, 'RELATION', rel=rel)
lmk_words, lmkw_prob, lmkw_ent = get_words(lmk_exp, 'LANDMARK-PHRASE', lmk=lmk)
sentence = ' '.join(rel_words + lmk_words)
if consistent:
meaning1 = m2s(lmk,rel)
# get the most likely meaning for the generated sentence
posteriors = get_sentence_posteriors(sentence)
meaning2 = max(posteriors, key=itemgetter(1))[0]
# is this what we are trying to say?
if meaning1 != meaning2:
continue
return sentence
def get_sentence_posteriors(sentence, iterations=1):
probs = []
meanings = []
# parse sentence with charniak and apply surgeries
print 'parsing ...'
modparse = get_modparse(sentence)
t = ParentedTree.parse(modparse)
print '\n%s\n' % t.pprint()
num_ancestors = count_lmk_phrases(t) - 1
for _ in xrange(iterations):
meaning = get_meaning(num_ancestors=num_ancestors)
lmk, rel = meaning
probs.append(get_tree_prob(t, *meaning))
meanings.append(m2s(lmk,rel))
print '.'
probs = np.array(probs) / sum(probs)
return uniquify_distribution(meanings, probs)
def accept_new_words( location, sentence, update_func='geometric', update_scale=10, num_meanings=10, printing=True ):
for _ in range(num_meanings):
print '.',
sys.stdout.flush()
(lmk, lmk_prob, lmk_ent), (rel, rel_prob, rel_ent) = get_meaning(loc=location)
old_meaning_prob, old_meaning_entropy, lrpc, tps = get_sentence_meaning_likelihood( sentence, lmk, rel, printing )
update = 10 * update_scale
# reward new words with old meaning
for lhs,rhs,parent,lmk,rel in lrpc:
# print 'Incrementing production - lhs: %s, rhs: %s, parent: %s' % (lhs,rhs,parent)
update_expansion_counts( update, lhs, rhs, parent, rel=rel,
lmk_class=(lmk.object_class if lmk else None),
lmk_ori_rels=get_lmk_ori_rels_str(lmk),
lmk_color=(lmk.color if lmk else None) )
for lhs,rhs,lmk,rel in tps:
# print 'Incrementing word - pos: %s, word: %s, lmk_class: %s' % (lhs, rhs, (lmk.object_class if lmk else None) )
update_word_counts( update, lhs, rhs, lmk_class=(lmk.object_class if lmk else None),
rel=rel,
lmk_ori_rels=get_lmk_ori_rels_str(lmk),
lmk_color=(lmk.color if lmk else None) )
print
# unpack row
xloc, yloc, sentence, parse, modparse = row
# convert variables to the right types
xloc = float(xloc)
yloc = float(yloc)
loc = (xloc, yloc)
parse = ParentedTree.parse(parse)
modparse = ParentedTree.parse(modparse)
# how many ancestors should the sampled landmark have?
num_ancestors = count_lmk_phrases(modparse) - 1
# sample `args.iterations` times for each sentence
for _ in xrange(args.iterations):
lmk, rel = get_meaning(loc, num_ancestors)
if args.verbose:
print "utterance:", repr(sentence)
print "location: %s" % repr(loc)
print "landmark: %s (%s)" % (lmk, lmk_id(lmk))
print "relation: %s" % rel_type(rel)
print "parse:"
print parse.pprint()
print "modparse:"
print modparse.pprint()
print "-" * 70
location = Location(x=xloc, y=yloc)
save_tree(modparse, location, rel, lmk)
Bigram.make_bigrams(location.words)
xloc = float(xloc)
yloc = float(yloc)
loc = (xloc, yloc)
parse = ParentedTree.parse(parse)
modparse = ParentedTree.parse(modparse)
# how many ancestors should the sampled landmark have?
num_ancestors = count_lmk_phrases(modparse) - 1
if num_ancestors == -1:
print 'Failed to parse %d [%s] [%s] [%s]' % (i, sentence, parse, modparse)
continue
# sample `args.iterations` times for each sentence
for _ in xrange(args.iterations):
lmk, rel = get_meaning(loc, num_ancestors)
lmk, _, _ = lmk
rel, _, _ = rel
assert(not isinstance(lmk, tuple))
assert(not isinstance(rel, tuple))
if args.verbose:
print 'utterance:', repr(sentence)
print 'location: %s' % repr(loc)
print 'landmark: %s (%s)' % (lmk, lmk_id(lmk))
print 'relation: %s' % rel_type(rel)
print 'parse:'
print parse.pprint()
print 'modparse:'
print modparse.pprint()
