def test_generate(self):
# test MultiPOSSwap transformation
gene = POSGenerator(
transformation_methods=["SwapMultiPOS"],
subpopulation_methods=[],
transformation_config={"SwapMultiPOS": [{
"treebank_tag": "NN"
}]})
for original_samples, trans_rst, trans_type in gene.generate(dataset):
self.assertEqual(2, len(original_samples))
self.assertEqual(2, len(trans_rst))
for index in range(2):
self.assertTrue(trans_rst[index].get_mask('x')[-1] == 2)
self.assertTrue(trans_rst[index].get_words('x')[-1] !=
original_samples[index].get_words('x')[-1])
# test PrefixSwap transformation
gene = POSGenerator(transformation_methods=['SwapPrefix'],
subpopulation_methods=[])
for original_samples, trans_rst, trans_type in gene.generate(dataset):
self.assertEqual(2, len(original_samples))
self.assertEqual(2, len(trans_rst))
for index in range(2):
self.assertTrue(trans_rst[index].get_mask('x')[-2] == 2)
self.assertTrue(trans_rst[index].get_words('x')[-2] !=
original_samples[index].get_words('x')[-2])
# test wrong transformation_methods
gene = POSGenerator(transformation_methods=["wrong_transform_method"],
subpopulation_methods=[])
self.assertRaises(ValueError, next, gene.generate(dataset))
gene = POSGenerator(transformation_methods=["AddSubtree"],
subpopulation_methods=[])
self.assertRaises(ValueError, next, gene.generate(dataset))
gene = POSGenerator(transformation_methods="OOV",
subpopulation_methods=[])
self.assertRaises(ValueError, next, gene.generate(dataset))
# test empty dataset
self.assertRaises(ValueError, next, gene.generate(Dataset('POS')))
# test empty sample
self.assertRaises(ValueError, next, gene.generate(special_dataset))
def test_generate(self):
test1 = CWSSample({'x': '', 'y': []})
test2 = CWSSample({'x': '~ ! @ # $ % ^ & * ( ) _ +', 'y': []})
dataset = Dataset('CWS')
dataset.load([test1, test2])
mode = [
'SwapName', 'CnSwapNum', 'Reduplication', 'CnMLM',
'SwapContraction', 'SwapVerb', 'SwapSyn'
]
gene = CWSGenerator(transformation_methods=mode,
subpopulation_methods=[])
for original_samples, trans_rst, trans_type in gene.generate(dataset):
self.assertTrue(len(original_samples) == 0)
self.assertTrue(len(trans_rst) == 0)
# test wrong transformation_methods
gene = CWSGenerator(transformation_methods=["wrong_transform_method"],
subpopulation_methods=[])
self.assertRaises(ValueError, next, gene.generate(dataset))
gene = CWSGenerator(transformation_methods=["AddSubtree"],
subpopulation_methods=[])
self.assertRaises(ValueError, next, gene.generate(dataset))
gene = CWSGenerator(transformation_methods="CnMLM",
subpopulation_methods=[])
self.assertRaises(ValueError, next, gene.generate(dataset))
sent1 = '周小明生产一万'
sent2 = '央视想朦胧'
dataset = Dataset(task='CWS')
dataset.load({
'x': [sent1, sent2],
'y': [['B', 'M', 'E', 'B', 'E', 'B', 'E'],
['B', 'E', 'S', 'B', 'E']]
})
gene = CWSGenerator(transformation_methods=mode,
subpopulation_methods=[])
for original_samples, trans_rst, trans_type in gene.generate(dataset):
self.assertTrue(len(original_samples) == len(trans_rst))
{'context': context, 'question': 'Which NFL team represented the '
'AFC at Super Bowl 50?',
'answers': [{"text": "Denver Broncos", "answer_start": 177},
{"text": "Denver Broncos", "answer_start": 177},
{"text": "Denver Broncos", "answer_start": 177}],
'title': "Super_Bowl_50", 'is_impossible': False})
sample2 = MRCSample(
{'context': " ", 'question': 'Which NFL team represented '
'the AFC at Super Bowl 50?',
'answers': [], 'title': "Super_Bowl_50", 'is_impossible': True})
sample3 = MRCSample(
{'context': "! @ # $ % ^ & * ( )",
'question': 'Which NFL team represented the AFC at Super Bowl 50?',
'answers': [], 'title': "Super_Bowl_50", 'is_impossible': True})
dataset = Dataset('MRC')
dataset.load(data_sample)
dataset.extend([sample2, sample3])
class TestMRCGenerator(unittest.TestCase):
def test_generate(self):
# test task transformation
# TODO, domain transformation addsentdiverse
transformation_methods = ["PerturbAnswer", "ModifyPos"]
gene = MRCGenerator(transformation_methods=transformation_methods,
subpopulation_methods=[])
for original_samples, trans_rst, trans_type in gene.generate(dataset):
self.assertEqual(1, len(trans_rst))
for index in range(len(original_samples)):
from TextFlint.generation_layer.generator.coref_generator import CorefGenerator
from TextFlint.input_layer.dataset import Dataset
import unittest
from ....data.coref_debug import CorefDebug
sample1 = CorefDebug.coref_sample1()
sample2 = CorefDebug.coref_sample2()
sample3 = CorefDebug.coref_sample3()
sample4 = CorefDebug.coref_sample4()
sample5 = CorefDebug.coref_sample5()
sample6 = CorefDebug.coref_sample6()
samples = [sample1, sample2, sample3, sample4, sample5, sample6]
dataset = Dataset("COREF")
dataset.load(samples)
class TestRndShuffle(unittest.TestCase):
def test_transform(self):
gene = CorefGenerator(transformation_methods=["RndShuffle"],
subpopulation_methods=[])
for original_samples, trans_rst, trans_type in gene.generate(dataset):
self.assertEqual(len(original_samples), len(trans_rst))
for so, st in zip(original_samples, trans_rst):
self.assertTrue(so.num_sentences() == st.num_sentences())
if __name__ == "__main__":
unittest.main()
import unittest
from TextFlint.input_layer.dataset import Dataset
from TextFlint.generation_layer.generator.ner_generator import NERGenerator
sample1 = {'x': 'Amy lives in a city , which is called NYK .',
'y': ['B-PER', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-LOC', 'O']}
sample2 = {'x': 'Jotion lives in Xian 105 kilometers away .',
'y': ['B-PER', 'O', 'O', 'B-LOC', 'O', 'O', 'O', 'O']}
sample3 = {'x': 'China rejects Syrians call to boycott Chinese lamb .',
'y': ['B-ORG', 'O', 'B-MISC', 'O', 'O', 'O', 'B-MISC', 'O', 'O']}
single_data_sample = [sample1]
data_samples = [sample1, sample2, sample3]
dataset = Dataset('NER')
single_dataset = Dataset('NER')
dataset.load(data_samples)
single_dataset.load(single_data_sample)
gene = NERGenerator()
class TestSpecialEntityTyposSwap(unittest.TestCase):
def test_generate(self):
# test task transformation
transformation_methods = ["SwapEnt", "EntTypos"]
gene = NERGenerator(transformation_methods=transformation_methods,
subpopulation_methods=[])
for original_samples, trans_rst, trans_type in gene.generate(dataset):
self.assertEqual(3, len(original_samples))
for index in range(len(original_samples)):
for ori_entity, trans_entity in \
import unittest
from TextFlint.input_layer.dataset import Dataset
from TextFlint.generation_layer.generator.sa_generator import SAGenerator
sample1 = {'x': 'Titanic is my favorite movie.',
'y': 'pos'}
sample2 = {'x': 'I don\'t like the actor Tim Hill', 'y': 'neg'}
sample3 = {'x': 'The leading actor is good.',
'y': 'pos'}
sample4 = {'x': '',
'y': 'pos'}
sample5 = {'x': '!@#$$%^&*()_+}{|":?><',
'y': 'pos'}
single_data_sample = [sample1]
data_samples = [sample1, sample2, sample3, sample4, sample5]
dataset = Dataset('SA')
single_dataset = Dataset('SA')
dataset.load(data_samples)
single_dataset.load(single_data_sample)
class TestSpecialEntityTyposSwap(unittest.TestCase):
def test_generate(self):
# test task transformation
transformation_methods = ["SwapSpecialEnt", "AddSum",
"DoubleDenial", "SwapNum"]
SA_config = {'AddSum': [{'entity_type': 'movie'},
{'entity_type': 'person'}],
'SwapSpecialEnt': [{'entity_type': 'movie'},
{'entity_type': 'person'}]}
'x': ['That', 'is', 'a', 'pretty', 'prefixed', 'survey'],
'y': ['DT', 'VBZ', 'DT', 'RB', 'JJ', 'NN']
}
sample2 = {
'x': ['That', 'is', 'a', 'prefixed', 'survey'],
'y': ['DT', 'VBZ', 'DT', 'JJ', 'NN']
}
sample3 = {'x': ['', '', ''], 'y': ['O', 'O', 'O']}
sample4 = {
'x': '! @ # $ % ^ & * ( )',
'y': ['O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O']
}
special_data_sample = [sample3, sample4]
data_samples = [sample1, sample2]
dataset = Dataset('POS')
dataset.load(data_samples)
special_dataset = Dataset('POS')
special_dataset.load(special_data_sample)
class TestPOSGenerate(unittest.TestCase):
def test_generate(self):
# test MultiPOSSwap transformation
gene = POSGenerator(
transformation_methods=["SwapMultiPOS"],
subpopulation_methods=[],
transformation_config={"SwapMultiPOS": [{
"treebank_tag": "NN"
}]})
for original_samples, trans_rst, trans_type in gene.generate(dataset):
}
sample3 = {
'hypothesis': 'There are two little boys smiling.',
'premise': 'Two little boys are smiling and laughing while one is '
'standing and one is in a bouncy seat',
'y': 'entailment'
}
sample4 = {
'hypothesis': '! @ # $ % ^ & * ( )',
'premise': '! @ # $ % ^ & * ( )',
'y': 'neutral'
}
data_samples = [sample1, sample2, sample3, sample4]
dataset = Dataset(task='NLI')
dataset.load(data_samples)
gene = NLIGenerator()
class TestNLIGenerator(unittest.TestCase):
def test_generate(self):
# test task transformation, ignore NliOverlap because it
# does't rely on the original data
transformation_methods = ["SwapAnt", "AddSent", "NumWord"]
gene = NLIGenerator(transformation_methods=transformation_methods,
subpopulation_methods=[])
for original_samples, trans_rst, trans_type in gene.generate(dataset):
for index in range(len(original_samples)):
logger.info(original_samples[index].dump())
import unittest
from TextFlint.input_layer.dataset import Dataset
from TextFlint.generation_layer.generator.dp_generator import DPGenerator, sample, sample_1
single_data_sample = [sample]
data_samples = [sample, sample_1]
dataset = Dataset('DP')
dataset.load(data_samples)
gene = DPGenerator()
class TestDPGenerator(unittest.TestCase):
def test_generate(self):
# test task transformation
gene = DPGenerator(transformation_methods=["DeleteSubTree"],
subpopulation_methods=[])
for original_samples, trans_rst, trans_type in gene.generate(dataset):
self.assertEqual(2, len(original_samples))
for original_sample, transformed_sample in \
zip(original_samples, trans_rst):
self.assertTrue(
len(original_sample.get_value('x')) != len(
transformed_sample.get_value('x')))
transformation_methods = ["DeleteSubTree", "Ocr"]
gene = DPGenerator(transformation_methods=transformation_methods,
subpopulation_methods=[])
for original_samples, trans_rst, trans_type in gene.generate(dataset):
for ori_sample, trans_sample in zip(original_samples, trans_rst):
self.assertTrue(ori_sample != trans_sample)
"sentence": "! @ # $ % ^ & * ( )",
"term_list": {
"35390182#756337#4_0": {
"id": "35390182#756337#4_0",
"polarity": "positive",
"term": "!",
"from": 0,
"to": 1,
"opinion_words": ["@"],
"opinion_position": [[2, 3]]
}
}
}
data_samples = [sample1, sample2, sample3]
dataset = Dataset('ABSA')
dataset.load(data_samples)
special_samples = [sample4, sample5]
special_dataset = Dataset('ABSA')
special_dataset.load(special_samples)
class TestABSAGenerator(unittest.TestCase):
def test_generate(self):
# test task transformation
transformation_methods = ['RevTgt', 'RevNon', 'AddDiff']
gene = ABSAGenerator(transformation_methods=transformation_methods,
subpopulation_methods=[],
dataset_config='restaurant')
'sentence2': 'Mr zhang has 20 students',
'y': '0'}
sample2 = {'sentence1': 'I like eating apples',
'sentence2': 'I love to eat apples',
'y': '1'}
sample3 = {'sentence1': 'There are two little boys smiling.',
'sentence2': 'Two little boys are smiling and laughing '
'while one is standing and one is in a bouncy seat',
'y': '0'}
sample4 = {'sentence1': '! @ # $ % ^ & * ( )',
'sentence2': '! @ # $ % ^ & * ( )',
'y': '0'}
data_samples = [sample1, sample2, sample3, sample4]
dataset = Dataset(task='SM')
dataset.load(data_samples)
gene = SMGenerator()
class TestSMGenerator(unittest.TestCase):
def test_generate(self):
# test task transformation, ignore SmOverlap because
# it does't rely on the original data
transformation_methods = ["SwapWord", "SwapNum"]
gene = SMGenerator(transformation_methods=transformation_methods,
subpopulation_methods=[])
for original_samples, trans_rst, trans_type in gene.generate(dataset):
for index in range(len(original_samples)):
# test whether the sample changed or not