def _prepare_data(self, seqs, slots, with_labels=True):
x = []
y_seq_id = []
y_time = []
y_labels = [[] for slot in slots]
for item in seqs:
x_vecs = []
for features in item['data']:
x_vec = np.zeros((len(self.vocab), ))
for ftr, val in features.iteritems():
x_vec[self.vocab[ftr]] = val
x_vecs.append(x_vec)
x.append(x_vecs)
labels = item['labels']
for label in labels:
y_seq_id.append(len(x) - 1)
y_time.append(label['time'])
for i, slot in enumerate(slots):
lbl_val = label['slots'][slot]
if lbl_val < 0:
lbl_val = len(self.slot_classes[slot]) + lbl_val
y_labels[i].append(lbl_val)
x_zero_pad = np.zeros((len(self.vocab), ))
x = padded(x, pad_by=[x_zero_pad]).transpose(1, 0, 2)
data = [x]
data.extend([y_seq_id, y_time])
if with_labels:
data.extend(y_labels)
return tuple(data)
def _prepare_data(self, seqs, slots, with_labels=True):
x = []
y_seq_id = []
y_time = []
y_labels = [[] for slot in slots]
for item in seqs:
x_vecs = []
for features in item['data']:
x_vec = np.zeros((len(self.vocab), ))
for ftr, val in features.iteritems():
x_vec[self.vocab[ftr]] = val
x_vecs.append(x_vec)
x.append(x_vecs)
labels = item['labels']
for label in labels:
y_seq_id.append(len(x) - 1)
y_time.append(label['time'])
for i, slot in enumerate(slots):
lbl_val = label['slots'][slot]
if lbl_val < 0:
lbl_val = len(self.slot_classes[slot]) + lbl_val
y_labels[i].append(lbl_val)
x_zero_pad = np.zeros((len(self.vocab), ))
x = padded(x, pad_by=[x_zero_pad]).transpose(1, 0, 2)
data = [x]
data.extend([y_seq_id, y_time])
if with_labels:
data.extend(y_labels)
return tuple(data)
def _prepare_data(self, seqs, slots, with_labels=True):
x = []
x_score = []
x_actor = []
y_seq_id = []
y_time = []
y_labels = [[] for slot in slots]
y_weights = []
for item in seqs:
x.append(item['data'])
x_score.append(item['data_score'])
x_actor.append(item['data_actor'])
labels = item['labels']
for label in labels:
y_seq_id.append(len(x) - 1)
y_time.append(label['time'])
for i, slot in enumerate(slots):
lbl_val = label['slots'][slot]
if lbl_val < 0:
lbl_val = len(self.slot_classes[slot]) + lbl_val
y_labels[i].append(lbl_val)
y_weights.append(label['score'])
x = padded(x, is_int=True).transpose(1, 0)
x_score = padded(x_score).transpose(1, 0)
x_actor = padded(x_actor, is_int=True).transpose(1, 0)
x_score = np.array(x_score, dtype=np.int32)[:,:]
y_weights = np.array(y_weights, dtype=np.float32)
y_token_labels_padding = self._prepare_y_token_labels_padding()
data = [x]
if self.x_include_score:
data.append(x_score)
data.extend([y_seq_id, y_time])
if with_labels:
data.append(y_weights)
data.extend(y_labels)
return tuple(data)
def _prepare_data(self, seqs, slots, with_labels=True):
x = []
x_score = []
x_actor = []
y_seq_id = []
y_time = []
y_labels = [[] for slot in slots]
y_weights = []
for item in seqs:
x.append(item['data'])
x_score.append(item['data_score'])
x_actor.append(item['data_actor'])
labels = item['labels']
for label in labels:
y_seq_id.append(len(x) - 1)
y_time.append(label['time'])
for i, slot in enumerate(slots):
lbl_val = label['slots'][slot]
if lbl_val < 0:
lbl_val = len(self.slot_classes[slot]) + lbl_val
y_labels[i].append(lbl_val)
y_weights.append(label['score'])
x = padded(x, is_int=True).transpose(1, 0)
x_score = padded(x_score).transpose(1, 0)
x_actor = padded(x_actor, is_int=True).transpose(1, 0)
x_score = np.array(x_score, dtype=np.int32)[:,:]
y_weights = np.array(y_weights, dtype=np.float32)
y_token_labels_padding = self._prepare_y_token_labels_padding()
data = [x]
if self.x_include_score:
data.append(x_score)
data.extend([y_seq_id, y_time])
if with_labels:
data.append(y_weights)
data.extend(y_labels)
return tuple(data)