def word_sim(self, word1, word2):
if (word1, word2) in self.sim_cache:
return self.sim_cache[(word1, word2)]
cand1 = self.get_spell_variations(word1)
cand2 = self.get_spell_variations(word2)
max_pair = (word1, word2)
max_sim = None
for c1 in cand1:
for c2 in cand2:
if c1 in self.model and c2 in self.model:
s = self.model.similarity(c1, c2)
# logging.info(
# u'Calling similarity: \t{0}\t{1}'.format(
# c1, c2).encode('utf8'))
if not max_sim or s > max_sim:
max_pair = (word1, word2)
max_sim = s
if max_sim is not None:
if max_sim > 0.1:
if self.wordnet_boost:
D = Wordnet.get_boost(max_pair[0], max_pair[1])
if D is not None:
max_sim += 0.5 * math.exp(-0.25 * D)
else:
max_sim = 0.0
if max_sim is not None and max_sim > 1.0:
max_sim = 1.0
self.sim_cache[(word1, word2)] = max_sim
self.sim_cache[(word2, word1)] = max_sim
return max_sim
def add_wordnet_senses(self, tokens):
for token in tokens:
if self.conf.getboolean('wordnet', 'enrich_with_senses'):
token['senses'] = Wordnet.get_senses(
token['token'],
self.conf.getint('wordnet', 'sense_threshold'))
else:
token['senses'] = set([token['token']])
def _featurize(self, w1, w2):
s1 = Wordnet.get_significant_synsets(w1)
s2 = Wordnet.get_significant_synsets(w2)
yield 'wordnet_hyp', float(Wordnet.is_hypernym(s1, s2))
yield 'wordnet_2-hyp', float(Wordnet.is_two_link_hypernym(s1, s2))
yield 'wordnet_deriv_rel', float(
Wordnet.is_derivationally_related(s1, s2))
yield 'wordnet_in_glosses', float(Wordnet.in_glosses(w1, w2, s1, s2))
def add_wordnet_senses(self, tokens):
for token in tokens:
if self.conf.getboolean('wordnet', 'enrich_with_senses'):
token['senses'] = Wordnet.get_senses(token['token'], self.conf.getint('wordnet', 'sense_threshold'))
else:
token['senses'] = set([token['token']])
def add_wordnet_senses(self, tokens):
for token in tokens:
if self.conf.getboolean("wordnet", "enrich_with_senses"):
token["senses"] = Wordnet.get_senses(token["token"], self.conf.getint("wordnet", "sense_threshold"))
else:
token["senses"] = set([token["token"]])
