def __init__(self, lang='en'):
try:
assertion_pickle = open('data/concept_assertion.pickle', 'rb')
self.concept_assertions = pickle.load(assertion_pickle)
assertion_pickle.close()
except:
self.concept_assertions = {}
self.nlp = simplenlp.get(lang)
self.stopwords = stopwords.words('english')
self.stopwords.extend(concept_stopwords)
def lookup_concept_from_nl(language, text):
"""
Look up a concept using any natural language text that represents it.
This function requires the :mod:`simplenlp` module
to normalize natural language text into a raw concept name.
"""
import simplenlp
nltools = simplenlp.get('en')
normalized = nltools.normalize(text)
return lookup_concept_raw(language, normalized)
def make_concept_uri(text, lang, disambiguation=None):
if lang == 'en':
from metanl import english
normalized, disambig = english.normalize_topic(text)
elif lang == 'ja':
from metanl import japanese
normalized, disambig = japanese.normalize(text), None
elif lang in ('pt', 'hu', 'nl', 'es'):
# languages where we know the stopword list
import simplenlp
nlp = simplenlp.get(lang)
disambig = None
normalized, disambig = nlp.normalize(text), None
else:
normalized = text
disambig = None
if disambiguation is not None:
disambig = disambiguation
if disambig is not None:
disambig = disambig.replace(' ', '_')
if disambig:
return '/c/%s/%s/%s' % (lang, normalized.replace(' ', '_'), disambig)
else:
return '/c/%s/%s' % (lang, normalized.replace(' ', '_'))
def make_concept_uri(text, lang, disambiguation=None):
if lang == 'en':
from metanl import english
normalized, disambig = english.normalize_topic(text)
elif lang == 'ja':
from metanl import japanese
normalized, disambig = japanese.normalize(text), None
elif lang in ('pt', 'hu', 'nl', 'es'):
# languages where we know the stopword list
import simplenlp
nlp = simplenlp.get(lang)
disambig = None
normalized, disambig = nlp.normalize(text), None
else:
normalized = text
disambig = None
if disambiguation is not None:
disambig = disambiguation
if disambig is not None:
disambig = disambig.replace(' ', '_')
if disambig:
return '/c/%s/%s/%s' % (lang, normalized.replace(' ', '_'), disambig)
else:
return '/c/%s/%s' % (lang, normalized.replace(' ', '_'))
import simplenlp
EN = simplenlp.get('en')
from conceptnet5.english_nlp import normalize
def check_line(line):
parts = line.strip().split()
norm = normalize(parts[0])
if norm != parts[1]:
print "Original: %s / WordNet: %s / conceptnet: %s" %\
(parts[0], parts[1], norm)
def compare_words():
for line in open('/Users/rspeer/nltk_data/corpora/wordnet/noun.exc'):
check_line(line)
for line in open('/Users/rspeer/nltk_data/corpora/wordnet/verb.exc'):
check_line(line)
if __name__ == '__main__':
compare_words()
# -*- coding: utf-8 -*-
import nltk
from nltk.corpus import wordnet
import simplenlp
import re
EN = simplenlp.get('en')
try:
morphy = wordnet.morphy
except LookupError:
nltk.download('wordnet')
morphy = wordnet.morphy
STOPWORDS = ['the', 'a', 'an']
EXCEPTIONS = {
# Avoid obsolete and obscure roots, the way lexicographers don't.
'wrought': 'wrought', # not 'work'
'media': 'media', # not 'medium'
'installed': 'install', # not 'instal'
'installing': 'install',# not 'instal'
'synapses': 'synapse', # not 'synapsis'
'soles': 'sole', # not 'sol'
'pubes': 'pube', # not 'pubis'
'dui': 'dui', # not 'duo'
'comics': 'comic', # WordNet's root for this will make you nerd-rage
'taxis': 'taxi', # not 'taxis'
'teeth': 'tooth', # not 'teeth'
# Avoid nouns that shadow more common verbs.
'am': 'be',
def put_raw_assertion_in_graph(raw):
try:
lang = raw.language_id
if raw.frame.goodness < 1: return
if lang.startswith('zh'): return
polarity = raw.frame.frequency.value
activity = raw.sentence.activity.name
if 'rubycommons' in activity: return
# build the assertion
raw_arg1 = GRAPH.get_or_create_concept(lang, raw.text1)
raw_arg2 = GRAPH.get_or_create_concept(lang, raw.text2)
frame_text = raw.frame.text
if polarity > 0:
frame_text = frame_text.replace('{%}', '')
else:
frame_text = frame_text.replace('{%}', 'not')
frame = GRAPH.get_or_create_frame(lang, frame_text)
raw_assertion = GRAPH.get_or_create_assertion(
frame,
[raw_arg1, raw_arg2],
{'dataset': 'conceptnet/4/'+lang, 'license': 'CC-By', 'normalized': False}
)
# create justification structure
creator = raw.sentence.creator.username
if creator == 'verbosity': return
creator_node = GRAPH.get_or_create_node(
u'/source/contributor/omcs/'+creator
)
activity_node = GRAPH.get_or_create_node(u'/source/activity/omcs/'+activity)
GRAPH.justify(OMCS, activity_node)
GRAPH.justify(OMCS, creator_node)
conjunction = GRAPH.get_or_create_conjunction(
[creator_node, activity_node]
)
GRAPH.justify(conjunction, raw_assertion)
# make the normalized version
if lang == 'en':
arg1 = GRAPH.get_or_create_concept('en', en_normalize(raw.text1))
arg2 = GRAPH.get_or_create_concept('en', en_normalize(raw.text2))
elif lang == 'ja':
arg1 = GRAPH.get_or_create_concept('ja', JA.normalize(raw.text1))
arg2 = GRAPH.get_or_create_concept('ja', JA.normalize(raw.text2))
else:
nlp = simplenlp.get(lang)
arg1 = GRAPH.get_or_create_concept(lang, nlp.normalize(raw.text1))
arg2 = GRAPH.get_or_create_concept(lang, nlp.normalize(raw.text2))
if polarity > 0:
relation = GRAPH.get_or_create_relation(raw.frame.relation.name)
else:
relation = GRAPH.get_or_create_relation('Not'+raw.frame.relation.name)
assertion = GRAPH.get_or_create_assertion(
relation, [arg1, arg2],
{'dataset': 'conceptnet/4/'+lang, 'license': 'CC-By', 'normalized': True}
)
for vote in raw.votes.all():
voter = GRAPH.get_or_create_node(
u'/source/contributor/omcs/'+vote.user.username
)
GRAPH.justify(OMCS, voter)
GRAPH.justify(voter, raw_assertion, weight=vote.vote)
GRAPH.derive_normalized(raw_assertion, assertion)
print assertion
except Exception:
import traceback
traceback.print_exc()
from csc_utils.batch import queryset_foreach
from conceptnet.models import Sentence, Assertion, RawAssertion
from conceptnet5.graph import JSONWriterGraph
from conceptnet5.english_nlp import normalize as en_normalize
import simplenlp
GRAPH = JSONWriterGraph('json_data/conceptnet')
OMCS = GRAPH.get_or_create_node('/source/site/omcs')
GRAPH.justify('/', OMCS)
JA = simplenlp.get('ja')
# monkey-patch
def answer_false(*args):
return False
JA.is_stopword_record = answer_false
def put_raw_assertion_in_graph(raw):
try:
lang = raw.language_id
if raw.frame.goodness < 1: return
if lang.startswith('zh'): return
polarity = raw.frame.frequency.value
activity = raw.sentence.activity.name
if 'rubycommons' in activity: return
# build the assertion
raw_arg1 = GRAPH.get_or_create_concept(lang, raw.text1)
raw_arg2 = GRAPH.get_or_create_concept(lang, raw.text2)
frame_text = raw.frame.text
if polarity > 0:
import simplenlp
from metanl import english
import math, random
from luminoso3.background_space import get_commonsense_assoc
from colorizer.color_data import make_lab_color_data, lab_to_rgb, rgb_to_hsv
from colorizer.colorvote import weighted_elect_samples
ENGLISH = simplenlp.get('en')
ASSOC = get_commonsense_assoc('en', 100)
COLORDATA = {}
origdata = make_lab_color_data()
def importance_factor(colorname):
imp = 10000 / math.sqrt(english.word_frequency(colorname.split()[0], 1000000))
return int(imp)
for key, values in origdata.items():
subset_values = random.sample(values,
min(len(values), int(math.ceil(importance_factor(key)*math.sqrt(len(values))))))
COLORDATA[key] = subset_values
def output_colors(labcolors):
return [lab_to_rgb(c) for c in sorted(labcolors)]
class IncrementalColorizer(object):
def __init__(self, ncolors):
self.ncolors = ncolors
self.colors = [(128,128,128)] * ncolors
def test_nai():
ja = simplenlp.get('ja')
test_sentence = u'いいえ、分かりませんでした。'
assert ja.normalize(test_sentence) == u'いいえ 分かる ない'
def test_normalize():
ja = simplenlp.get('ja')
test_sentence = u'これはテストですが、大丈夫です。'
assert ja.normalize_list(test_sentence) == [u'テスト', u'大丈夫']
assert ja.normalize(test_sentence) == u'テスト 大丈夫'
def test_chinese():
zh = simplenlp.get('zh-Hant')
railway = u"迪士尼线"
assert zh.normalize(railway) == railway
def test_normalize():
ja = simplenlp.get('ja')
test_sentence = u'これはテストですが、大丈夫です。'
assert ja.normalize_list(test_sentence) == [u'テスト', u'大丈夫']
assert ja.normalize(test_sentence) == u'テスト 大丈夫'
def test_nai():
ja = simplenlp.get('ja')
test_sentence = u'いいえ、分かりませんでした。'
assert ja.normalize(test_sentence) == u'いいえ 分かる ない'
def test_chinese():
zh = simplenlp.get('zh-Hant')
railway = u"迪士尼线"
assert zh.normalize(railway) == railway
