def test_compute_similar_words_from_word(self) -> None:
file_path = str(
Path(__file__).resolve().parent / "samples" / "sample.word2vec.txt"
)
word_embeds = WordEmbedding(file_path)
word = "てすと"
expected = [("サンプル", 0.7506074093675397)]
actual = word_embeds.compute_similar_words_from_word(word)
self.assertEqual(actual, expected)
def test_property_shape(self) -> None:
file_path = str(
Path(__file__).resolve().parent / "samples" / "sample.word2vec.txt"
)
word_embeds = WordEmbedding(file_path)
self.assertEqual(len(word_embeds), 2)
self.assertEqual(word_embeds.dim, 10)
self.assertEqual(word_embeds.shape, (2, 10))
def test_error_unknown_words(self) -> None:
file_path = str(
Path(__file__).resolve().parent / "samples" / "sample.word2vec.txt"
)
word_embeds = WordEmbedding(file_path)
words = ["てすと", "unk"]
with self.assertRaises(ValueError, msg=f"unknown word: 'unk'"):
word_embeds(words)
def test_extract_word_embeddings(self) -> None:
file_path = str(
Path(__file__).resolve().parent / "samples" / "sample.word2vec.txt"
)
word_embeds = WordEmbedding(file_path)
words = ["てすと", "サンプル"]
actual = word_embeds(words)
expected = np.array(
[
[
0.31882,
0.89289,
0.90071,
0.45753,
0.37083,
0.64955,
0.34075,
0.70048,
0.89085,
0.13621,
],
[
0.79375,
0.44464,
0.07644,
0.35242,
0.03996,
0.68827,
0.97103,
0.77324,
0.72781,
0.69158,
],
]
)
np.testing.assert_almost_equal(actual, expected)
word = "てすと"
actual = word_embeds(word)
expected = np.array(
[
0.31882,
0.89289,
0.90071,
0.45753,
0.37083,
0.64955,
0.34075,
0.70048,
0.89085,
0.13621,
]
)
np.testing.assert_almost_equal(actual, expected)
def test_compute_cosine_similarity(self) -> None:
file_path = str(
Path(__file__).resolve().parent / "samples" / "sample.word2vec.txt"
)
word_embeds = WordEmbedding(file_path)
embed = np.array(
[
0.29902,
0.90019,
0.89964,
0.50753,
0.38001,
0.59495,
0.29175,
0.69909,
0.90185,
0.09687,
]
)
expected = np.array([0.9987114080207757, 0.7286216119815097])
actual = word_embeds.compute_cosine_similarity(embed)
np.testing.assert_almost_equal(actual, expected)
def test_compute_similar_words_from_vec(self) -> None:
file_path = str(
Path(__file__).resolve().parent / "samples" / "sample.word2vec.txt"
)
word_embeds = WordEmbedding(file_path)
embed = np.array(
[
0.29902,
0.90019,
0.89964,
0.50753,
0.38001,
0.59495,
0.29175,
0.69909,
0.90185,
0.09687,
]
)
expected = [("てすと", 0.9987114080207757), ("サンプル", 0.7286216119815097)]
actual = word_embeds.compute_similar_words_from_vec(embed)
self.assertEqual(actual, expected)
def test_property_vocab(self) -> None:
file_path = str(
Path(__file__).resolve().parent / "samples" / "sample.word2vec.txt"
)
word_embeds = WordEmbedding(file_path)
self.assertSetEqual(word_embeds.vocab, {"てすと", "サンプル"})
self.assertEqual(word_embeds.to_word(1), "サンプル")
self.assertEqual(word_embeds.to_index("てすと"), 0)
self.assertTrue(word_embeds.is_known("てすと"))
self.assertFalse(word_embeds.is_known("test"))
def test_load_glove_format_file(self) -> None:
file_path = str(
Path(__file__).resolve().parent / "samples" / "sample.glove.txt"
)
word_embeds = WordEmbedding(file_path)
self.assertIsInstance(word_embeds, WordEmbedding)