-
Notifications
You must be signed in to change notification settings - Fork 479
/
entity_lstm.py
490 lines (413 loc) · 24.8 KB
/
entity_lstm.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
import os
import pickle
import re
import time
import tensorflow as tf
from neuroner import utils
from neuroner import utils_tf
from neuroner import utils_nlp
def bidirectional_LSTM(inputs, hidden_state_dimension, initializer,
sequence_length=None, output_sequence=True):
"""
"""
with tf.variable_scope("bidirectional_LSTM"):
if sequence_length == None:
batch_size = 1
sequence_length = tf.shape(inputs)[1]
sequence_length = tf.expand_dims(sequence_length, axis=0,
name='sequence_length')
else:
batch_size = tf.shape(sequence_length)[0]
lstm_cell = {}
initial_state = {}
for direction in ["forward", "backward"]:
with tf.variable_scope(direction):
# LSTM cell
lstm_cell[direction] = tf.contrib.rnn.CoupledInputForgetGateLSTMCell(hidden_state_dimension,
forget_bias=1.0, initializer=initializer, state_is_tuple=True)
# initial state: http:https://stackoverflow.com/questions/38441589/tensorflow-rnn-initial-state
initial_cell_state = tf.get_variable("initial_cell_state", shape=[1, hidden_state_dimension],
dtype=tf.float32, initializer=initializer)
initial_output_state = tf.get_variable("initial_output_state", shape=[1,
hidden_state_dimension], dtype=tf.float32, initializer=initializer)
c_states = tf.tile(initial_cell_state, tf.stack([batch_size, 1]))
h_states = tf.tile(initial_output_state, tf.stack([batch_size, 1]))
initial_state[direction] = tf.contrib.rnn.LSTMStateTuple(c_states,
h_states)
# sequence_length must be provided for tf.nn.bidirectional_dynamic_rnn due to internal bug
outputs, final_states = tf.nn.bidirectional_dynamic_rnn(lstm_cell["forward"],
lstm_cell["backward"],inputs = inputs, dtype=tf.float32, sequence_length=sequence_length,
initial_state_fw=initial_state["forward"], initial_state_bw=initial_state["backward"])
if output_sequence == True:
outputs_forward, outputs_backward = outputs
output = tf.concat([outputs_forward, outputs_backward], axis=2, name='output_sequence')
else:
# # max pooling
# outputs_forward, outputs_backward = outputs
# output = tf.concat([outputs_forward, outputs_backward], axis=2, name='output_sequence')
# output = tf.reduce_max(output, axis=1, name='output')
# # last pooling
final_states_forward, final_states_backward = final_states
output = tf.concat([final_states_forward[1], final_states_backward[1]], axis=1, name='output')
return output
class EntityLSTM(object):
"""
An LSTM architecture for named entity recognition.
Uses a character embedding layer followed by an LSTM to generate vector
representation from characters for each token.
Then the character vector is concatenated with token embedding vector,
which is input to another LSTM followed by a CRF layer.
"""
def __init__(self, dataset, parameters):
self.verbose = False
# Placeholders for input, output and dropout
self.input_token_indices = tf.placeholder(tf.int32, [None],
name="input_token_indices")
self.input_label_indices_vector = tf.placeholder(tf.float32, [None,
dataset.number_of_classes], name="input_label_indices_vector")
self.input_label_indices_flat = tf.placeholder(tf.int32, [None],
name="input_label_indices_flat")
self.input_token_character_indices = tf.placeholder(tf.int32, [None, None],
name="input_token_indices")
self.input_token_lengths = tf.placeholder(tf.int32, [None],
name="input_token_lengths")
self.dropout_keep_prob = tf.placeholder(tf.float32,
name="dropout_keep_prob")
# Internal parameters
initializer = tf.contrib.layers.xavier_initializer()
if parameters['use_character_lstm']:
# Character-level LSTM
# Idea: reshape so that we have a tensor [number_of_token, max_token_length,
# token_embeddings_size], which we pass to the LSTM
# Character embedding layer
with tf.variable_scope("character_embedding"):
self.character_embedding_weights = tf.get_variable(
"character_embedding_weights",
shape=[dataset.alphabet_size, parameters['character_embedding_dimension']],
initializer=initializer)
embedded_characters = tf.nn.embedding_lookup(self.character_embedding_weights,
self.input_token_character_indices, name='embedded_characters')
if self.verbose:
print("embedded_characters: {0}".format(embedded_characters))
utils_tf.variable_summaries(self.character_embedding_weights)
# Character LSTM layer
with tf.variable_scope('character_lstm') as vs:
character_lstm_output = bidirectional_LSTM(embedded_characters,
parameters['character_lstm_hidden_state_dimension'], initializer,
sequence_length=self.input_token_lengths, output_sequence=False)
self.character_lstm_variables = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES,
scope=vs.name)
# Token embedding layer
with tf.variable_scope("token_embedding"):
self.token_embedding_weights = tf.get_variable(
"token_embedding_weights",
shape=[dataset.vocabulary_size, parameters['token_embedding_dimension']],
initializer=initializer,
trainable=not parameters['freeze_token_embeddings'])
embedded_tokens = tf.nn.embedding_lookup(self.token_embedding_weights,
self.input_token_indices)
utils_tf.variable_summaries(self.token_embedding_weights)
# Concatenate character LSTM outputs and token embeddings
if parameters['use_character_lstm']:
with tf.variable_scope("concatenate_token_and_character_vectors"):
if self.verbose:
print('embedded_tokens: {0}'.format(embedded_tokens))
token_lstm_input = tf.concat([character_lstm_output, embedded_tokens],
axis=1, name='token_lstm_input')
if self.verbose:
print("token_lstm_input: {0}".format(token_lstm_input))
else:
token_lstm_input = embedded_tokens
# Add dropout
with tf.variable_scope("dropout"):
token_lstm_input_drop = tf.nn.dropout(token_lstm_input, self.dropout_keep_prob,
name='token_lstm_input_drop')
if self.verbose:
print("token_lstm_input_drop: {0}".format(token_lstm_input_drop))
# https://www.tensorflow.org/api_guides/python/contrib.rnn
# Prepare data shape to match `rnn` function requirements
# Current data input shape: (batch_size, n_steps, n_input)
# Required shape: 'n_steps' tensors list of shape (batch_size, n_input)
token_lstm_input_drop_expanded = tf.expand_dims(token_lstm_input_drop,
axis=0, name='token_lstm_input_drop_expanded')
if self.verbose:
print("token_lstm_input_drop_expanded: {0}".format(token_lstm_input_drop_expanded))
# Token LSTM layer
with tf.variable_scope('token_lstm') as vs:
token_lstm_output = bidirectional_LSTM(token_lstm_input_drop_expanded,
parameters['token_lstm_hidden_state_dimension'], initializer, output_sequence=True)
token_lstm_output_squeezed = tf.squeeze(token_lstm_output, axis=0)
self.token_lstm_variables = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES,
scope=vs.name)
# Needed only if Bidirectional LSTM is used for token level
with tf.variable_scope("feedforward_after_lstm") as vs:
W = tf.get_variable(
"W",
shape=[2 * parameters['token_lstm_hidden_state_dimension'],
parameters['token_lstm_hidden_state_dimension']],
initializer=initializer)
b = tf.Variable(tf.constant(0.0, shape=[parameters['token_lstm_hidden_state_dimension']]),
name="bias")
outputs = tf.nn.xw_plus_b(token_lstm_output_squeezed, W, b, name="output_before_tanh")
outputs = tf.nn.tanh(outputs, name="output_after_tanh")
utils_tf.variable_summaries(W)
utils_tf.variable_summaries(b)
self.token_lstm_variables += tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, scope=vs.name)
with tf.variable_scope("feedforward_before_crf") as vs:
W = tf.get_variable(
"W",
shape=[parameters['token_lstm_hidden_state_dimension'],
dataset.number_of_classes],
initializer=initializer)
b = tf.Variable(tf.constant(0.0, shape=[dataset.number_of_classes]),
name="bias")
scores = tf.nn.xw_plus_b(outputs, W, b, name="scores")
self.unary_scores = scores
self.predictions = tf.argmax(self.unary_scores, 1, name="predictions")
utils_tf.variable_summaries(W)
utils_tf.variable_summaries(b)
self.feedforward_variables = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES,
scope=vs.name)
# CRF layer
if parameters['use_crf']:
with tf.variable_scope("crf") as vs:
# Add start and end tokens
small_score = -1000.0
large_score = 0.0
sequence_length = tf.shape(self.unary_scores)[0]
unary_scores_with_start_and_end = tf.concat([self.unary_scores,
tf.tile( tf.constant(small_score, shape=[1, 2]) , [sequence_length, 1])], 1)
start_unary_scores = [[small_score] * dataset.number_of_classes + [large_score, small_score]]
end_unary_scores = [[small_score] * dataset.number_of_classes + [small_score, large_score]]
self.unary_scores = tf.concat([start_unary_scores, unary_scores_with_start_and_end,
end_unary_scores], 0)
start_index = dataset.number_of_classes
end_index = dataset.number_of_classes + 1
input_label_indices_flat_with_start_and_end = tf.concat([tf.constant(start_index,
shape=[1]), self.input_label_indices_flat, tf.constant(end_index,
shape=[1]) ], 0)
# Apply CRF layer
sequence_length = tf.shape(self.unary_scores)[0]
sequence_lengths = tf.expand_dims(sequence_length, axis=0,
name='sequence_lengths')
unary_scores_expanded = tf.expand_dims(self.unary_scores, axis=0,
name='unary_scores_expanded')
input_label_indices_flat_batch = tf.expand_dims(input_label_indices_flat_with_start_and_end,
axis=0, name='input_label_indices_flat_batch')
if self.verbose:
print('unary_scores_expanded: {0}'.format(unary_scores_expanded))
print('input_label_indices_flat_batch: {0}'.format(input_label_indices_flat_batch))
print("sequence_lengths: {0}".format(sequence_lengths))
# https://github.com/tensorflow/tensorflow/tree/master/tensorflow/contrib/crf
# Compute the log-likelihood of the gold sequences and keep the
# transition params for inference at test time.
self.transition_parameters=tf.get_variable(
"transitions",
shape=[dataset.number_of_classes+2, dataset.number_of_classes+2],
initializer=initializer)
utils_tf.variable_summaries(self.transition_parameters)
log_likelihood, _ = tf.contrib.crf.crf_log_likelihood(
unary_scores_expanded, input_label_indices_flat_batch, sequence_lengths,
transition_params=self.transition_parameters)
self.loss = tf.reduce_mean(-log_likelihood, name='cross_entropy_mean_loss')
self.accuracy = tf.constant(1)
self.crf_variables = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES,
scope=vs.name)
# Do not use CRF layer
else:
with tf.variable_scope("crf") as vs:
self.transition_parameters = tf.get_variable(
"transitions",
shape=[dataset.number_of_classes+2, dataset.number_of_classes+2],
initializer=initializer)
utils_tf.variable_summaries(self.transition_parameters)
self.crf_variables = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES,
scope=vs.name)
# Calculate mean cross-entropy loss
with tf.variable_scope("loss"):
losses = tf.nn.softmax_cross_entropy_with_logits(logits=self.unary_scores,
labels=self.input_label_indices_vector, name='softmax')
self.loss = tf.reduce_mean(losses, name='cross_entropy_mean_loss')
with tf.variable_scope("accuracy"):
correct_predictions = tf.equal(self.predictions,
tf.argmax(self.input_label_indices_vector, 1))
self.accuracy = tf.reduce_mean(tf.cast(correct_predictions, 'float'),
name='accuracy')
self.define_training_procedure(parameters)
self.summary_op = tf.summary.merge_all()
# defaults to saving all variables
self.saver = tf.train.Saver(max_to_keep=parameters['maximum_number_of_epochs'])
def define_training_procedure(self, parameters):
"""
Define training procedure
"""
self.global_step = tf.Variable(0, name="global_step", trainable=False)
if parameters['optimizer'] == 'adam':
self.optimizer = tf.train.AdamOptimizer(parameters['learning_rate'])
elif parameters['optimizer'] == 'sgd':
self.optimizer = tf.train.GradientDescentOptimizer(parameters['learning_rate'])
elif parameters['optimizer'] == 'adadelta':
self.optimizer = tf.train.AdadeltaOptimizer(parameters['learning_rate'])
else:
raise ValueError('The lr_method parameter must be either adadelta, adam or sgd.')
grads_and_vars = self.optimizer.compute_gradients(self.loss)
if parameters['gradient_clipping_value']:
grads_and_vars = [(tf.clip_by_value(grad, -parameters['gradient_clipping_value'],
parameters['gradient_clipping_value']), var) for grad, var in grads_and_vars]
# By defining a global_step variable and passing it to the optimizer
# we allow TensorFlow handle the counting of training steps for us.
# The global step will be automatically incremented by one every time
# you execute train_op.
self.train_op = self.optimizer.apply_gradients(grads_and_vars, global_step=self.global_step)
def load_pretrained_token_embeddings(self, sess, dataset, parameters,
token_to_vector=None):
"""
"""
if parameters['token_pretrained_embedding_filepath'] == '':
return
# Load embeddings
start_time = time.time()
print('Load token embeddings... ', end='', flush=True)
if token_to_vector == None:
token_to_vector = utils_nlp.load_pretrained_token_embeddings(parameters)
initial_weights = sess.run(self.token_embedding_weights.read_value())
number_of_loaded_word_vectors = 0
number_of_token_original_case_found = 0
number_of_token_lowercase_found = 0
number_of_token_digits_replaced_with_zeros_found = 0
number_of_token_lowercase_and_digits_replaced_with_zeros_found = 0
for token in dataset.token_to_index.keys():
if token in token_to_vector.keys():
initial_weights[dataset.token_to_index[token]] = token_to_vector[token]
number_of_token_original_case_found += 1
elif parameters['check_for_lowercase'] and token.lower() in token_to_vector.keys():
initial_weights[dataset.token_to_index[token]] = token_to_vector[token.lower()]
number_of_token_lowercase_found += 1
elif parameters['check_for_digits_replaced_with_zeros'] and re.sub(r'\d',
'0', token) in token_to_vector.keys():
initial_weights[dataset.token_to_index[token]] = token_to_vector[re.sub(r'\d',
'0', token)]
number_of_token_digits_replaced_with_zeros_found += 1
elif parameters['check_for_lowercase'] and parameters['check_for_digits_replaced_with_zeros'] \
and re.sub('\d', '0', token.lower()) in token_to_vector.keys():
initial_weights[dataset.token_to_index[token]] = token_to_vector[re.sub(r'\d',
'0', token.lower())]
number_of_token_lowercase_and_digits_replaced_with_zeros_found += 1
else:
continue
number_of_loaded_word_vectors += 1
elapsed_time = time.time() - start_time
print('done ({0:.2f} seconds)'.format(elapsed_time))
print("number_of_token_original_case_found: {0}".format(number_of_token_original_case_found))
print("number_of_token_lowercase_found: {0}".format(number_of_token_lowercase_found))
print("number_of_token_digits_replaced_with_zeros_found: {0}".format(number_of_token_digits_replaced_with_zeros_found))
print("number_of_token_lowercase_and_digits_replaced_with_zeros_found: {0}".format(number_of_token_lowercase_and_digits_replaced_with_zeros_found))
print('number_of_loaded_word_vectors: {0}'.format(number_of_loaded_word_vectors))
print("dataset.vocabulary_size: {0}".format(dataset.vocabulary_size))
sess.run(self.token_embedding_weights.assign(initial_weights))
def load_embeddings_from_pretrained_model(self, sess, dataset, pretraining_dataset,
pretrained_embedding_weights, embedding_type='token'):
"""
"""
if embedding_type == 'token':
embedding_weights = self.token_embedding_weights
index_to_string = dataset.index_to_token
pretraining_string_to_index = pretraining_dataset.token_to_index
elif embedding_type == 'character':
embedding_weights = self.character_embedding_weights
index_to_string = dataset.index_to_character
pretraining_string_to_index = pretraining_dataset.character_to_index
# Load embeddings
start_time = time.time()
print('Load {0} embeddings from pretrained model... '.format(embedding_type),
end='', flush=True)
initial_weights = sess.run(embedding_weights.read_value())
if embedding_type == 'token':
initial_weights[dataset.UNK_TOKEN_INDEX] = pretrained_embedding_weights[pretraining_dataset.UNK_TOKEN_INDEX]
elif embedding_type == 'character':
initial_weights[dataset.PADDING_CHARACTER_INDEX] = pretrained_embedding_weights[pretraining_dataset.PADDING_CHARACTER_INDEX]
number_of_loaded_vectors = 1
for index, string in index_to_string.items():
if index == dataset.UNK_TOKEN_INDEX:
continue
if string in pretraining_string_to_index.keys():
initial_weights[index] = pretrained_embedding_weights[pretraining_string_to_index[string]]
number_of_loaded_vectors += 1
elapsed_time = time.time() - start_time
print('done ({0:.2f} seconds)'.format(elapsed_time))
print("number_of_loaded_vectors: {0}".format(number_of_loaded_vectors))
if embedding_type == 'token':
print("dataset.vocabulary_size: {0}".format(dataset.vocabulary_size))
elif embedding_type == 'character':
print("dataset.alphabet_size: {0}".format(dataset.alphabet_size))
sess.run(embedding_weights.assign(initial_weights))
def restore_from_pretrained_model(self, parameters, dataset, sess,
token_to_vector=None):
"""
"""
try:
pretraining_dataset = pickle.load(open(os.path.join(parameters['pretrained_model_folder'],
'dataset.pickle'), 'rb'))
except:
pretraining_dataset = utils.renamed_load(open(os.path.join(parameters['pretrained_model_folder'],
'dataset.pickle'), 'rb'))
pretrained_model_checkpoint_filepath = os.path.join(parameters['pretrained_model_folder'],
'model.ckpt')
# Assert that the label sets are the same
# Test set should have the same label set as the pretrained dataset
assert pretraining_dataset.index_to_label == dataset.index_to_label
# If the token and character mappings are exactly the same
if pretraining_dataset.index_to_token == dataset.index_to_token and \
pretraining_dataset.index_to_character == dataset.index_to_character:
# Restore the pretrained model
# Works only when the dimensions of tensor variables are matched.
self.saver.restore(sess, pretrained_model_checkpoint_filepath)
# If the token and character mappings are different between the pretrained model
# and the current model
else:
# Resize the token and character embedding weights to match them with
# the pretrained model (required in order to restore the pretrained model)
utils_tf.resize_tensor_variable(sess, self.character_embedding_weights,
[pretraining_dataset.alphabet_size, parameters['character_embedding_dimension']])
utils_tf.resize_tensor_variable(sess, self.token_embedding_weights,
[pretraining_dataset.vocabulary_size, parameters['token_embedding_dimension']])
# Restore the pretrained model
# Works only when the dimensions of tensor variables are matched.
self.saver.restore(sess, pretrained_model_checkpoint_filepath)
# Get pretrained embeddings
character_embedding_weights, token_embedding_weights = sess.run([self.character_embedding_weights,
self.token_embedding_weights])
# Restore the sizes of token and character embedding weights
utils_tf.resize_tensor_variable(sess, self.character_embedding_weights,
[dataset.alphabet_size, parameters['character_embedding_dimension']])
utils_tf.resize_tensor_variable(sess, self.token_embedding_weights,
[dataset.vocabulary_size, parameters['token_embedding_dimension']])
# Re-initialize the token and character embedding weights
sess.run(tf.variables_initializer([self.character_embedding_weights,
self.token_embedding_weights]))
# Load embedding weights from pretrained token embeddings first
self.load_pretrained_token_embeddings(sess, dataset, parameters,
token_to_vector=token_to_vector)
# Load embedding weights from pretrained model
self.load_embeddings_from_pretrained_model(sess, dataset, pretraining_dataset,
token_embedding_weights, embedding_type='token')
self.load_embeddings_from_pretrained_model(sess, dataset, pretraining_dataset,
character_embedding_weights, embedding_type='character')
del pretraining_dataset
del character_embedding_weights
del token_embedding_weights
# Get transition parameters
transition_params_trained = sess.run(self.transition_parameters)
if not parameters['reload_character_embeddings']:
sess.run(tf.variables_initializer([self.character_embedding_weights]))
if not parameters['reload_character_lstm']:
sess.run(tf.variables_initializer(self.character_lstm_variables))
if not parameters['reload_token_embeddings']:
sess.run(tf.variables_initializer([self.token_embedding_weights]))
if not parameters['reload_token_lstm']:
sess.run(tf.variables_initializer(self.token_lstm_variables))
if not parameters['reload_feedforward']:
sess.run(tf.variables_initializer(self.feedforward_variables))
if not parameters['reload_crf']:
sess.run(tf.variables_initializer(self.crf_variables))
return transition_params_trained