def test_set_vocab(self):
model_file = tempfile.NamedTemporaryFile()
Word2Vec.train(model_file, self.training_data, size=100, window=5, negative=5, min_count=0, workers=4)
w2v = Word2Vec(model_file.name)
w2v.set_vocab(["誰", "私", "彼"])
self.assertEqual(len(w2v.syn0norm_in_vocab), 3)
model_file.close()
def test_most_similar_word1(self):
model_file = tempfile.NamedTemporaryFile()
Word2Vec.train(model_file, self.training_data, size=100, window=5, negative=5, min_count=0, workers=4)
w2v = Word2Vec(model_file.name, topn=5)
with self.assertRaisesRegex(AssertionError, r"^You need to call set_vocab first$"):
_ = w2v.most_similar_word("人間")
model_file.close()
def test_similarity(self):
model_file = tempfile.NamedTemporaryFile()
Word2Vec.train(model_file, self.training_data, size=100, window=5, negative=5, min_count=0, workers=4)
w2v = Word2Vec(model_file.name, topn=5)
sim = w2v.similarity("THIS_WORD_IS_NOT_IN_THE_TRAINING_DATA", "彼")
self.assertAlmostEqual(sim, 0.0)
sim = w2v.similarity("彼", "私")
self.assertNotAlmostEquals(sim, 0.0)
model_file.close()
def test_most_similar_word2(self):
model_file = tempfile.NamedTemporaryFile()
Word2Vec.train(model_file, self.training_data, size=100, window=5, negative=5, min_count=0, workers=4)
w2v = Word2Vec(model_file.name, topn=5)
w2v.set_vocab(["誰", "私", "彼"])
most_sim_words = list(map(itemgetter(0), w2v.most_similar_word("人間")))
self.assertEqual(len(most_sim_words), 3)
self.assertIn("誰", most_sim_words)
self.assertIn("私", most_sim_words)
self.assertIn("彼", most_sim_words)
model_file.close()
def _load_tgt_embedding(self):
if self._should_load_tgt_embedding():
return Word2Vec(model_path=self.setting.tgt_embedding)
else:
return None
def _load_src_embedding(self):
if self._should_load_src_embedding():
return Word2Vec(model_path=self.setting.src_embedding, topn=self.setting.src_embedding_topn)
else:
return None
def _load_src_embedding(self):
return Word2Vec(model_path=self.setting.src_embedding, topn=self.setting.src_embedding_topn)
def test_train(self):
model_file = tempfile.NamedTemporaryFile()
Word2Vec.train(model_file, self.training_data, size=100, window=5, negative=5, min_count=0, workers=4)
model_file.close()