def get_topics(self, word):
self.candidates = []
with WordLoader() as word_loader, SenseLoader() as sense_loader:
words = word_loader.load_words_with_lemma(word)
if not words:
return
for sense in sense_loader.load_senses_with_synset(words[0]):
self.find_synlinks_recursively(sense, 'hype')
return list(set(self.candidates)) # uniq
def find_synlinks_recursively(self, senses, link, lang='jpn', _depth=0):
if senses is None:
return
with SynlinkLoader() as synlink_loader, WordLoader(
) as word_loader, SenseLoader() as sense_loader:
word = word_loader.load_word_with_wordid(senses.wordid).lemma
synlink = synlink_loader.load_synlinks_with_sense_and_link(
senses, link)
print(f"{word} ", end='')
if (word in self.categories):
self.candidates.append(word)
if not synlink:
return
sense = sense_loader.load_sense_with_synset(synlink[0].synset2)
self.find_synlinks_recursively(sense, link)
def get_hypos(self, word):
self.hypos = []
with SynlinkLoader() as synlink_loader, WordLoader(
) as word_loader, SenseLoader() as sense_loader:
words = word_loader.load_words_with_lemma(word)
if not words:
return
for sense in sense_loader.load_senses_with_synset(words[0]):
for synlink in synlink_loader.load_synlinks_with_sense_and_link(
sense, 'hypo'):
hypo = sense_loader.load_sense_with_synset(synlink.synset2)
if hypo:
self.hypos.append(
word_loader.load_word_with_wordid(
hypo.wordid).lemma)
return self.hypos # uniq
def get_senses(word_to_check, synset_data):
if word_to_check in GET_SENSE_CALCULATED:
return GET_SENSE_CALCULATED[word_to_check]
words = WordLoader().load_words_with_lemma(word_to_check)
senses = []
senses_vecs = []
for word in words:
#print(word)
senses_this_ittr = SenseLoader().load_senses_with_synset(word)
#senses_this_ittr = SenseLoader().load_senses_with_synset_plus(word)
senses += senses_this_ittr
senses_vecs += get_vecs_from_senses(synset_data, senses_this_ittr)
#print(senses)
'''
if len(senses) > 200:
print("This word: ", word_to_check)
GET_SENSE_CALCULATED[word_to_check] = ([],[])
return ([],[])
'''
GET_SENSE_CALCULATED[word_to_check] = (senses, senses_vecs)
return (senses, senses_vecs)
def print_synlinks_recursively(senses, link, lang='jpn', _depth=0):
for sense in senses:
with SynlinkLoader() as synlink_loader:
synlinks = synlink_loader.load_synlinks_with_sense_and_link(
sense, link)
if synlinks:
with WordLoader() as word_loader:
with SynsetLoader() as synset_loader:
print(''.join([
' ' * 2 * _depth,
word_loader.load_word_with_wordid(sense.wordid).lemma,
' ',
synset_loader.load_synset_with_synset(sense.synset).name]))
_senses = []
for synLink in synlinks:
with SenseLoader() as sense_loader:
sense = sense_loader.load_sense_with_synset(
synLink.synset2, lang)
if sense:
_senses.append(sense)
print_synlinks_recursively(_senses, link, lang, _depth + 1)
for synLink in synlinks:
with SenseLoader() as sense_loader:
sense = sense_loader.load_sense_with_synset(
synLink.synset2, lang)
if sense:
_senses.append(sense)
print_synlinks_recursively(_senses, link, lang, _depth + 1)
if __name__ == '__main__':
if len(sys.argv) >= 3:
with WordLoader() as word_loader:
words = word_loader.load_words_with_lemma(sys.argv[1])
if words:
with SenseLoader() as sense_loader:
senses = sense_loader.load_senses_with_synset(words[0])
if len(sys.argv) >= 3:
link = sys.argv[2]
else:
link = 'hypo'
if len(sys.argv) == 4:
lang = sys.argv[3]
else:
lang = 'jpn'
print_synlinks_recursively(senses, link, lang)
sys.exit()
if count == 0:
return total
return total / count
if __name__ == '__main__':
word2vec_dic = load_word2vec_dic()
synset_data = load_synset_data()
total = 0
correct = 0
for wordid in range(155288, 249121):
if wordid % 1000 == 0:
print(wordid)
word = WordLoader().load_word_with_wordid(wordid)
lemma_senses = SenseLoader().load_senses_with_synset(word)
lemma_senses_vecs = get_vecs_from_senses(synset_data, lemma_senses)
chosen_sense_index = Random().randint(0, len(lemma_senses) - 1)
chosen_synset_id = lemma_senses[chosen_sense_index].synset
synset = SynsetLoader().load_synset_with_synset(chosen_synset_id)
similar_synset_ids = get_similar_synset_ids(synset)
similar_synset_avg_vector = average_synset_vectors(
similar_synset_ids, synset_data)
sense_ranking = get_closest_sense(similar_synset_avg_vector,
lemma_senses, lemma_senses_vecs,
euclidean)
if sense_ranking[0][0].synset == chosen_synset_id: #correct
correct += 1