def test_out_of_vocabulary(self):
vocabulary = Vocabulary(["a a a b b OOV c"], WordTokenizer(True), out_of_vocabulary="OOV")
self.assertEqual(1, vocabulary.index("OOV"))
self.assertEqual(1, vocabulary.index("z"))
self.assertEqual(2, vocabulary.index("a"))
self.assertEqual(3, vocabulary.index("b"))
self.assertEqual(4, vocabulary.index("c"))
def test_out_of_vocabulary(self):
vocabulary = Vocabulary(["a a a b b OOV c"],
WordTokenizer(True),
out_of_vocabulary="OOV")
self.assertEqual(1, vocabulary.index("OOV"))
self.assertEqual(1, vocabulary.index("z"))
self.assertEqual(2, vocabulary.index("a"))
self.assertEqual(3, vocabulary.index("b"))
self.assertEqual(4, vocabulary.index("c"))
def test_vocabulary_serialization(self):
vocabulary_name = os.path.join(self.directory, "vocabulary.pkl")
vocabulary = Vocabulary(["the quick brown fox jumped over the lazy dog"], WordTokenizer(True))
self._serialize(vocabulary_name, vocabulary)
self.assertTrue(os.path.isfile(vocabulary_name))
deserialized_vocabulary = self._deserialize(vocabulary_name)
self.assertEqual(vocabulary, deserialized_vocabulary)
s = "The quick black fox"
np.testing.assert_equal(vocabulary.index_string(s), deserialized_vocabulary.index_string(s))
def test_vocabulary_serialization(self):
vocabulary_name = os.path.join(self.directory, "vocabulary.pkl")
vocabulary = Vocabulary(
["the quick brown fox jumped over the lazy dog"],
WordTokenizer(True))
self._serialize(vocabulary_name, vocabulary)
self.assertTrue(os.path.isfile(vocabulary_name))
deserialized_vocabulary = self._deserialize(vocabulary_name)
self.assertEqual(vocabulary, deserialized_vocabulary)
s = "The quick black fox"
np.testing.assert_equal(vocabulary.index_string(s),
deserialized_vocabulary.index_string(s))
def test_invalid_partition(self):
self.assertRaises(
ValueError, PartitionedData.from_text, {
"train": ["red red red", "blue blue green"],
"validate": ["red blue blue orange"],
"test": ["green green red black"]
}, ["bogus"], Vocabulary.factory(WordTokenizer(True)))
def create_partitioned_data():
return PartitionedData.from_text(
{
"train": ["blue blue green", "red red red"],
"test": ["green green red black"],
"validate": ["red blue blue orange"]
}, ["train"], Vocabulary.factory(WordTokenizer(True)))
def test_limited_vocabulary(self):
v = Vocabulary(["to be or not to be"],
WordTokenizer(True),
max_vocabulary=2)
self.assertEqual({"to", "be"}, set(v.type_to_index.keys()))
self.assertEqual(2, len(v))
v = Vocabulary(["hamlet hamlet hamlet to be or not to be"],
WordTokenizer(True),
min_frequency=2)
self.assertEqual({"to", "be", "hamlet"}, set(v.type_to_index.keys()))
self.assertEqual(3, len(v))
v = Vocabulary(["hamlet hamlet hamlet to be or not to be"],
WordTokenizer(True),
max_vocabulary=2,
min_frequency=2)
self.assertEqual({"be", "hamlet"}, set(v.type_to_index.keys()))
self.assertEqual(2, len(v))
def test_full_vocabulary(self):
v = Vocabulary(["the quick brown fox jumped over the lazy dog"],
WordTokenizer(True))
self.assertEqual(
"Vocabulary, size 8: None:1 the:2 brown:3 dog:4 fox:5 ...", str(v))
self.assertEqual(
{"the", "quick", "brown", "fox", "jumped", "over", "lazy", "dog"},
set(v.type_to_index.keys()))
self.assertEqual(8, len(v))
def create_data_set(args):
if args.vocabulary_partitions is None:
args.vocabulary_partitions = args.partitions.keys()
tokenizer = {"whitespace": WhitespaceTokenizer,
"word": WordTokenizer,
"character": CharacterTokenizer}[args.tokenizer](args.case_normalized)
factory = Vocabulary.factory(tokenizer,
min_frequency=args.min_frequency, max_vocabulary=args.max_vocabulary,
out_of_vocabulary=args.out_of_vocabulary)
partitions = {}
for partition, filenames in args.partitions.items():
partitions[partition] = [open(filename).read() for filename in filenames]
partitioned_data = PartitionedData.from_text(partitions, args.vocabulary_partitions, factory)
partitioned_data.serialize(args.directory)
logger.info("Created %s in %s" % (partitioned_data, args.directory))
def create_data_set(args):
if args.vocabulary_partitions is None:
args.vocabulary_partitions = args.partitions.keys()
tokenizer = {
"whitespace": WhitespaceTokenizer,
"word": WordTokenizer,
"character": CharacterTokenizer
}[args.tokenizer](args.case_normalized)
factory = Vocabulary.factory(tokenizer,
min_frequency=args.min_frequency,
max_vocabulary=args.max_vocabulary,
out_of_vocabulary=args.out_of_vocabulary)
partitions = {}
for partition, filenames in args.partitions.items():
partitions[partition] = [
open(filename).read() for filename in filenames
]
partitioned_data = PartitionedData.from_text(partitions,
args.vocabulary_partitions,
factory)
partitioned_data.serialize(args.directory)
logger.info("Created %s in %s" % (partitioned_data, args.directory))
def test_invalid_index(self):
document = "the quick brown fox jumped over the lazy dog"
vocabulary = Vocabulary([document], WordTokenizer(True))
self.assertRaises(ValueError, vocabulary.type, 0)
self.assertRaises(ValueError, vocabulary.type, -1)
def test_vocabulary_factory(self):
factory = Vocabulary.factory(WordTokenizer(True), max_vocabulary=2)
self.assertEqual(
Vocabulary(["to be or not to be"],
WordTokenizer(True),
max_vocabulary=2), factory(["to be or not to be"]))
def test_index_tokens(self):
document = "the quick brown fox jumped over the lazy dog"
vocabulary = Vocabulary([document], WordTokenizer(True))
np.testing.assert_equal(np.array([2, 9, 3, 5, 6, 8, 2, 7, 4]), vocabulary.index_string(document))
def test_vocabulary_factory(self):
factory = Vocabulary.factory(WordTokenizer(True), max_vocabulary=2)
self.assertEqual(Vocabulary(["to be or not to be"], WordTokenizer(True), max_vocabulary=2),
factory(["to be or not to be"]))
def test_index_tokens(self):
document = "the quick brown fox jumped over the lazy dog"
vocabulary = Vocabulary([document], WordTokenizer(True))
np.testing.assert_equal(np.array([2, 9, 3, 5, 6, 8, 2, 7, 4]),
vocabulary.index_string(document))
def create_partitioned_data():
return PartitionedData.from_text({
"train": ["blue blue green", "red red red"],
"test": ["green green red black"],
"validate": ["red blue blue orange"]
}, ["train"], Vocabulary.factory(WordTokenizer(True)))
def test_invalid_partition(self):
self.assertRaises(ValueError, PartitionedData.from_text, {
"train": ["red red red", "blue blue green"],
"validate": ["red blue blue orange"],
"test": ["green green red black"]
}, ["bogus"], Vocabulary.factory(WordTokenizer(True)))