def test_n_output_audio(self):
audio, sampling_rate = AudioLoader.load_audio(self.sample_wav_file)
flows = [
naf.Sequential(
[naa.CropAug(sampling_rate=sampling_rate),
naa.LoudnessAug()]),
naf.Sometimes(
[naa.CropAug(sampling_rate=sampling_rate),
naa.LoudnessAug()],
pipeline_p=0.9),
naf.Sequential([
naf.Sequential([
naa.CropAug(sampling_rate=sampling_rate),
naa.LoudnessAug()
]),
naf.Sometimes([
naa.CropAug(sampling_rate=sampling_rate),
naa.LoudnessAug()
],
pipeline_p=0.9)
])
]
for flow in flows:
augmented_audios = flow.augment(audio, n=3)
self.assertGreater(len(augmented_audios), 1)
for augmented_audio in augmented_audios:
self.assertFalse(np.array_equal(audio, augmented_audio))
self.assertLess(0, len(flows))
def test_n_output_spectrogram(self):
mel_spectrogram = AudioLoader.load_mel_spectrogram(
self.sample_wav_file, n_mels=128)
#
flows = [
naf.Sequential([nas.FrequencyMaskingAug(),
nas.TimeMaskingAug()]),
naf.Sometimes([nas.FrequencyMaskingAug(),
nas.TimeMaskingAug()],
pipeline_p=0.9),
naf.Sequential([
naf.Sequential(
[nas.FrequencyMaskingAug(),
nas.TimeMaskingAug()]),
naf.Sometimes(
[nas.FrequencyMaskingAug(),
nas.TimeMaskingAug()],
pipeline_p=0.9)
])
]
for flow in flows:
augmented_mel_spectrograms = flow.augment(mel_spectrogram, n=3)
self.assertGreater(len(augmented_mel_spectrograms), 1)
for augmented_mel_spectrogram in augmented_mel_spectrograms:
self.assertFalse(
np.array_equal(mel_spectrogram, augmented_mel_spectrogram))
self.assertLess(0, len(flows))
def test_multiple_actions(self):
texts = [
'The quick brown fox jumps over the lazy dog',
'Zology raku123456 fasdasd asd4123414 1234584'
]
flows = [
naf.Sequential([
naf.Sometimes([
nac.RandomCharAug(action=Action.INSERT),
nac.RandomCharAug(action=Action.DELETE)
],
pipeline_p=0.5),
naf.Sequential([
nac.OcrAug(),
nac.QwertyAug(aug_min=1),
nac.RandomCharAug(action=Action.SUBSTITUTE,
aug_min=1,
aug_char_p=0.6,
aug_word_p=0.6)
],
name='Sub_Seq')
]),
naf.Sometimes([
naf.Sometimes([
nac.RandomCharAug(action=Action.INSERT),
nac.RandomCharAug(action=Action.DELETE)
]),
naf.Sequential([
nac.OcrAug(),
nac.QwertyAug(aug_min=1),
nac.RandomCharAug(action=Action.SUBSTITUTE,
aug_min=1,
aug_char_p=0.6,
aug_word_p=0.6)
])
],
pipeline_p=0.5)
]
# Since prob may be low and causing do not perform data augmentation. Retry 5 times
for flow in flows:
at_least_one_not_equal = False
for _ in range(0, 5):
for text in texts:
self.assertLess(0, len(text))
augmented_text = flow.augment(text)
if text != augmented_text:
at_least_one_not_equal = True
self.assertLess(0, len(text))
if at_least_one_not_equal:
break
self.assertTrue(at_least_one_not_equal)
self.assertLess(0, len(flows))
self.assertLess(0, len(texts))
def test_multiple_actions(self):
texts = [
'The quick brown fox jumps over the lazy dog',
'Zology raku123456 fasdasd asd4123414 1234584'
]
flows = [
naf.Sequential([
naf.Sometimes([
nac.RandomCharAug(action="insert"),
nac.RandomCharAug(action="delete")
],
pipeline_p=0.9),
naf.Sequential(
[
# nac.OcrAug(), nac.QwertyAug(aug_min=1),
nac.RandomCharAug(action="substitute",
aug_char_min=1,
aug_char_p=0.6,
aug_word_p=0.6)
],
name='Sub_Seq')
]),
naf.Sometimes([
naf.Sometimes([
nac.RandomCharAug(action="insert"),
nac.RandomCharAug(action="delete")
]),
naf.Sequential([
nac.OcrAug(),
nac.KeyboardAug(aug_char_min=1),
nac.RandomCharAug(action="substitute",
aug_char_min=1,
aug_char_p=0.6,
aug_word_p=0.6)
])
],
pipeline_p=0.9)
]
# Since prob may be low and causing do not perform data augmentation. Retry 5 times
for flow in flows:
for text in texts:
at_least_one_not_equal = False
for _ in range(5):
augmented_text = flow.augment(text, n=1)
if text != augmented_text:
at_least_one_not_equal = True
break
self.assertTrue(at_least_one_not_equal)
self.assertLess(0, len(flows))
self.assertLess(0, len(texts))
def augmentation(text, insert=False, substitute=False, swap=True, delete=True):
augs = []
if insert:
aug = naw.WordEmbsAug(
model_type='word2vec',
model_path=
'/media/jionie/my_disk/Kaggle/Tweet/model/word2vec/GoogleNews-vectors-negative300.bin',
action="insert")
augs.append(aug)
if substitute:
aug_sub = naw.SynonymAug(aug_src='wordnet')
augs.append(aug_sub)
if swap:
aug_swap = naw.RandomWordAug(action="swap")
augs.append(aug_swap)
if delete:
aug_del = naw.RandomWordAug()
augs.append(aug_del)
aug = naf.Sometimes(augs, aug_p=0.5, pipeline_p=0.5)
# print("before aug:", text)
text = aug.augment(text, n=1)
# print("after aug:", text)
return text
def test_n_output_without_augmentation(self):
texts = [
'AAAAAAAAAAA AAAAAAAAAAAAAA'
]
flows = [
naf.Sequential([
nac.OcrAug(),
nac.OcrAug()
]),
naf.Sometimes([
nac.RandomCharAug(),
nac.RandomCharAug()
], pipeline_p=0.00001)
]
for flow in flows:
for text in texts:
for _ in range(5):
augmented_texts = flow.augment(text, n=3)
all_not_equal = False
for augmented_text in augmented_texts:
if augmented_text != text:
all_not_equal = True
break
if all_not_equal:
break
self.assertFalse(all_not_equal)
self.assertLess(0, len(flows))
self.assertLess(0, len(texts))
def train_eval_dataset(dataset: pd.DataFrame,lang="ita",expansion=10):
nltk.download('averaged_perceptron_tagger')
nltk.download('wordnet')
nltk.download('omw')
flow = naf.Sometimes([naw.SynonymAug(lang=lang, aug_min=10),naw.RandomWordAug("swap"),naw.RandomWordAug("delete"),nac.KeyboardAug()])
train_afert_exp=[]
dev_after_exp=[]
for idx, row in dataset.iterrows():
logging.info("[{}/{}] {}".format(idx, len(dataset), row["question"]))
new_text = [new for new in flow.augment(row["question"], n=expansion)]
train_afert_exp.append({"label": row["question_id"], "text": row["question"]})
th=int(len(new_text)*0.8)
for text in new_text[:th]:
train_afert_exp.append({"label": row["question_id"], "text": text})
for text in new_text[th:]:
dev_after_exp.append({"label": row["question_id"], "text": text})
train=train_afert_exp
dev=dev_after_exp
train = pd.DataFrame(train).sample(frac=1.0)
dev = pd.DataFrame(dev).sample(frac=1.0)
return train, dev
def test_augment_detail(self):
text = 'The quick brown fox jumps over the lazy dog'
flows = [
naf.Sequential([
naf.Sometimes([
nac.RandomCharAug(action="insert"),
nac.RandomCharAug(action="delete")
],
pipeline_p=0.5),
naf.Sequential([
nac.RandomCharAug(action="substitute",
aug_char_min=1,
aug_char_p=0.6,
aug_word_p=0.6)
],
name='Sub_Seq')
],
include_detail=True),
naf.Sometimes([
naf.Sometimes([
nac.RandomCharAug(action="insert"),
nac.RandomCharAug(action="delete")
]),
naf.Sequential([
nac.OcrAug(),
nac.KeyboardAug(aug_char_min=1),
nac.RandomCharAug(action="substitute",
aug_char_min=1,
aug_char_p=0.6,
aug_word_p=0.6)
])
],
pipeline_p=1,
include_detail=True)
]
for flow in flows:
augmented_text, augment_details = flow.augment(text)
self.assertNotEqual(text, augmented_text)
self.assertGreater(len(augment_details), 0)
for augment_detail in augment_details:
self.assertGreater(augment_detail['orig_start_pos'], -1)
self.assertGreater(augment_detail['new_start_pos'], -1)
self.assertGreater(augment_detail['change_seq'], 0)
self.assertIn(augment_detail['action'], Action.getall())
def test_multi_thread(self):
text = 'The quick brown fox jumps over the lazy dog'
n = 3
w2v_model_path = os.path.join(os.environ["MODEL_DIR"], 'word',
'word_embs',
'GoogleNews-vectors-negative300.bin')
flows = [
naf.Sequential([
naf.Sequential([
nac.OcrAug(),
naw.WordEmbsAug(model_type='word2vec',
model_path=w2v_model_path)
]),
naf.Sequential([
nac.RandomCharAug(),
]),
naw.ContextualWordEmbsAug(model_path='xlnet-base-cased',
action="substitute",
temperature=0.7,
device='cpu')
]),
naf.Sometimes([
naf.Sequential([
nac.OcrAug(),
nac.RandomCharAug(),
]),
naf.Sometimes([
naw.WordEmbsAug(model_type='word2vec',
model_path=w2v_model_path)
],
pipeline_p=0.999),
naw.ContextualWordEmbsAug(model_path='xlnet-base-cased',
action="substitute",
temperature=0.7,
device='cpu')
],
pipeline_p=0.9999)
]
for num_thread in [1, 3]:
for flow in flows:
augmented_data = flow.augment(text, n=n, num_thread=num_thread)
self.assertEqual(len(augmented_data), n)
def parse(config: dict) -> list:
augmentations = []
for key, value in config.items():
au = AUGMENTATIONS.get(key, None)
if au is None:
raise KeyError(f"No augmentation named: {key}\n"
f"Available augmentations: {AUGMENTATIONS.keys()}")
aug = au(**value) if value is not None else au()
augmentations.append(aug)
return naf.Sometimes(augmentations)
def nlpaug(word):
aug = naf.Sometimes([
nac.OcrAug(),
nac.KeyboardAug(),
nac.RandomCharAug(action="insert"),
nac.RandomCharAug(action="substitute"),
nac.RandomCharAug(action="swap"),
nac.RandomCharAug(action="delete"),
naw.SpellingAug(),
])
word = aug.augment(word)
return word
def test_n_output(self):
texts = [
'The quick brown fox jumps over the lazy dog',
'Zology raku123456 fasdasd asd4123414 1234584',
'AAAAAAAAAAA AAAAAAAAAAAAAA'
]
flows = [
naf.Sequential(
[nac.RandomCharAug(action=Action.INSERT),
naw.RandomWordAug()]),
naf.Sometimes([
nac.RandomCharAug(action=Action.INSERT),
nac.RandomCharAug(action=Action.DELETE)
],
pipeline_p=0.9),
naf.Sequential([
naf.Sequential([
nac.RandomCharAug(action=Action.INSERT),
naw.RandomWordAug()
]),
naf.Sometimes([
nac.RandomCharAug(action=Action.INSERT),
nac.RandomCharAug(action=Action.DELETE)
],
pipeline_p=0.9)
])
]
for flow in flows:
for text in texts:
augmented_texts = flow.augment(text, n=3)
self.assertGreater(len(augmented_texts), 1)
for augmented_text in augmented_texts:
self.assertNotEqual(augmented_text, text)
self.assertLess(0, len(flows))
self.assertLess(0, len(texts))
def augmentation(self,
text,
insert=False,
substitute=False,
swap=True,
delete=True):
augs = []
if insert:
# aug = naw.ContextualWordEmbsAug(
# model_path=self.model_type, action="insert", device='cuda')
# wget https://s3.amazonaws.com/dl4j-distribution/GoogleNews-vectors-negative300.bin.gz
aug = naw.WordEmbsAug(
model_type='word2vec',
model_path=
'/C:/Users/admin/Documents/Nitin/mycodes/kaggle_google_quest_qna/data/helpers/word2vec/GoogleNews-vectors-negative300.bin',
action="insert")
augs.append(aug)
if substitute:
# aug = naw.ContextualWordEmbsAug(
# model_path=self.model_type, action="substitute", device='cuda')
# aug = naw.WordEmbsAug(
# model_type='word2vec', model_path='/media/jionie/my_disk/Kaggle/Google_Quest_Answer/model/word2vec/GoogleNews-vectors-negative300.bin',
# action="substitute")
aug_sub = naw.SynonymAug(aug_src='wordnet')
augs.append(aug_sub)
# text = aug.augment(text)
if swap:
aug_swap = naw.RandomWordAug(action="swap")
augs.append(aug_swap)
# text = aug.augment(text)
if delete:
aug_del = naw.RandomWordAug()
augs.append(aug_del)
# text = aug.augment(text)
aug = naf.Sometimes(augs, aug_p=0.5, pipeline_p=0.5)
# print("before aug:", text)
text = aug.augment(text, n=1)
# print("after aug:", text)
return text
def test_n_output_without_augmentation(self):
texts = ['AAAAAAAAAAA AAAAAAAAAAAAAA']
flows = [
naf.Sequential([nac.OcrAug(), nac.OcrAug()]),
naf.Sometimes(
[nac.RandomCharAug(), nac.RandomCharAug()], pipeline_p=0.00001)
]
for flow in flows:
for text in texts:
at_least_one_equal = False
for _ in range(5):
augmented_texts = flow.augment(text, n=3)
if len(augmented_texts
) == 1 and augmented_texts[0] == text:
at_least_one_equal = True
break
self.assertTrue(at_least_one_equal)
self.assertLess(0, len(flows))
self.assertLess(0, len(texts))
def test_single_action(self):
texts = [
'The quick brown fox jumps over the lazy dog',
'Zology raku123456 fasdasd asd4123414 1234584 s@#'
]
# Since prob may be low and causing do not perform data augmentation. Retry 5 times
at_least_one_not_equal = False
for _ in range(0, 5):
flow = naf.Sometimes([nac.RandomCharAug(action=Action.INSERT)], pipeline_p=0.6)
for text in texts:
augmented_text = flow.augment(text)
if text != augmented_text:
at_least_one_not_equal = True
self.assertLess(0, len(text))
if at_least_one_not_equal:
break
self.assertTrue(at_least_one_not_equal)
self.assertLess(0, len(texts))
def test_dry_run(self):
seq = naf.Sometimes()
results = seq.augment([])
self.assertEqual(0, len(results))