def infer(self, instance, *args, **kwargs):
instance = instance.strip().lower()
if not all([ord(c) < 128 for c in instance]):
print('Non ASCII symbols in the string, returning empty slots', file=sys.stderr)
return {}
tokens = tokenize_reg(instance)
if len(tokens) > 0:
return self.predict_slots(tokens)
else:
return {}
def __call__(self, batch, *args, **kwargs):
if isinstance(batch[0], str):
batch = [tokenize_reg(instance.strip()) for instance in batch]
slots = [{}] * len(batch)
m = [i for i, v in enumerate(batch) if v]
if m:
batch = [batch[i] for i in m]
tags_batch = self._ner_network.predict_for_token_batch(batch)
for i, tokens, tags in zip(m, batch, tags_batch):
slots[i] = self.predict_slots(tokens, tags)
return slots
def __call__(self, batch, *args, **kwargs):
if isinstance(batch[0], str):
batch = [tokenize_reg(instance.strip()) for instance in batch]
slots = [{}] * len(batch)
m = [i for i, v in enumerate(batch) if v]
if m:
batch = [batch[i] for i in m]
# tags_batch = self._ner_network.predict_for_token_batch(batch)
# batch example: [['is', 'there', 'anything', 'else']]
for i, tokens in zip(m, batch):
# tokens are['is', 'there', 'anything', 'else']
slots[i] = self._predict_slots(tokens, self.threshold)
# slots[i] example {'food': 'steakhouse'}
# slots we want, example : [{'pricerange': 'moderate', 'area': 'south'}]
return slots
def __call__(self, batch, *args, **kwargs):
if isinstance(batch[0], str):
batch = [tokenize_reg(instance.strip()) for instance in batch]
slots = [{}] * len(batch)
m = [i for i, v in enumerate(batch) if v]
if m:
batch = [batch[i] for i in m]
# tags_batch = self._ner_network.predict_for_token_batch(batch)
# batch example: [['is', 'there', 'anything', 'else']]
for i, tokens in zip(m, batch):
# tokens are['is', 'there', 'anything', 'else']
slots_values_lists = self._predict_slots(tokens)
if self.return_all:
slots[i] = dict(slots_values_lists)
else:
slots[i] = {slot: val_list[0] for slot, val_list in slots_values_lists.items()}
# slots[i] example {'food': 'steakhouse'}
# slots we want, example : [{'pricerange': 'moderate', 'area': 'south'}]
return slots
def __call__(self, batch, *args, **kwargs):
if isinstance(batch[0], str):
batch = [tokenize_reg(utterance) for utterance in batch]
return self._net.predict_on_batch(batch)