SentenceSimilarityGet_Based_Simas_LSTM/siamese_network_2.py at master · lqfarmer/SentenceSimilarityGet_Based_Simas_LSTM · GitHub

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
import tensorflow as tf
import numpy as np

class SiameseLSTM(object):
    """
    A LSTM based deep Siamese network for text similarity.
    Uses an character embedding layer, followed by a biLSTM and Energy Loss layer.
    """

    def BiRNN_FeatureExtract(self, x, dropout, scope, embedding_size, sequence_length):
        n_input=embedding_size
        n_steps=sequence_length
        n_hidden=240
        n_layers=3
        # Prepare data shape to match `bidirectional_rnn` function requirements
        # Current data input shape: (batch_size, n_steps, n_input) (?, seq_len, embedding_size)
        # Required shape: 'n_steps' tensors list of shape (batch_size, n_input)
        # Permuting batch_size and n_steps
        x = tf.transpose(x, [1, 0, 2])
        # Reshape to (n_steps*batch_size, n_input)
        x = tf.reshape(x, [-1, n_input])
        # Split to get a list of 'n_steps' tensors of shape (batch_size, n_input)
        #x = tf.split(0, n_steps, x)
        x = tf.split(x, n_steps, 0)
        print(x)
        # Define lstm cells with tensorflow
        # Forward direction cell
        with tf.name_scope("fwF"+scope),tf.variable_scope("fwF"+scope):
            print(tf.get_variable_scope().name)
            fw_cell = tf.contrib.rnn.LSTMCell(n_hidden, forget_bias=1.0, state_is_tuple=True)
#            fw_cell = tf.contrib.rnn.GRUCell(n_hidden,activation=tf.tanh)
            lstm_fw_cell = tf.contrib.rnn.DropoutWrapper(fw_cell,output_keep_prob=dropout)
            lstm_fw_cell_m=tf.contrib.rnn.MultiRNNCell([lstm_fw_cell]*n_layers, state_is_tuple=True)
        # Backward direction cell
        with tf.name_scope("bwF"+scope),tf.variable_scope("bwF"+scope):
            print(tf.get_variable_scope().name)
            bw_cell = tf.contrib.rnn.LSTMCell(n_hidden, forget_bias=1.0, state_is_tuple=True)
#            bw_cell = tf.contrib.rnn.GRUCell(n_hidden, activation=tf.tanh)
            lstm_bw_cell = tf.contrib.rnn.DropoutWrapper(bw_cell,output_keep_prob=dropout)
            lstm_bw_cell_m = tf.contrib.rnn.MultiRNNCell([lstm_bw_cell]*n_layers, state_is_tuple=True)
        # Get lstm cell output
        with tf.name_scope("bwF"+scope),tf.variable_scope("bwF"+scope):
            outputs, _, _ = tf.contrib.rnn.static_bidirectional_rnn(lstm_fw_cell_m, lstm_bw_cell_m, x, dtype=tf.float32)

        # create a dense layer and get final output
        output_temp = tf.layers.batch_normalization(tf.reduce_sum(outputs,0))
        outputL2=tf.nn.l2_normalize(output_temp,1)
        with tf.name_scope("dwF"+scope),tf.variable_scope("denselayer_wF"+scope),tf.variable_scope("denselayer_bF"+scope),tf.variable_scope("denselayer_outputF"+scope):
            output = tf.nn.xw_plus_b(output_temp,
                             tf.get_variable("denselayer_wF", [2 * n_hidden, n_hidden]),
                             tf.get_variable("denselayer_bF", [n_hidden]))
            output = tf.nn.elu(output, "densepayer_outputF")
        output = tf.layers.batch_normalization(output)
        with tf.name_scope("dw2F"+scope),tf.variable_scope("denselayer_w2F"+scope),tf.variable_scope("denselayer_b2F"+scope),tf.variable_scope("denselayer_output2F"+scope):
            output2 = tf.nn.xw_plus_b(output,
                             tf.get_variable("denselayer_w2F", [n_hidden, sequence_length]),
                             tf.get_variable("denselayer_b2F", [sequence_length]))
            output2 = tf.nn.elu(output2, "densepayer_output2F")
        output2 = tf.layers.batch_normalization(output2)
        return output2

    def BiRNN_AttentionMechanism(self, x, dropout, scope, embedding_size, sequence_length):
        n_input=embedding_size
        n_steps=sequence_length
        n_hidden=100
        n_layers=1
        # Prepare data shape to match `bidirectional_rnn` function requirements
        # Current data input shape: (batch_size, n_steps, n_input) (?, seq_len, embedding_size)
        # Required shape: 'n_steps' tensors list of shape (batch_size, n_input)
        # Permuting batch_size and n_steps
        x = tf.transpose(x, [1, 0, 2])
        # Reshape to (n_steps*batch_size, n_input)
        x = tf.reshape(x, [-1, n_input])
        # Split to get a list of 'n_steps' tensors of shape (batch_size, n_input)
        #x = tf.split(0, n_steps, x)
        x = tf.split(x, n_steps, 0)
        print(x)
        # Define lstm cells with tensorflow
        # Forward direction cell
        with tf.name_scope("fwA"+scope),tf.variable_scope("fwA"+scope):
            print(tf.get_variable_scope().name)
            fw_cell = tf.contrib.rnn.LSTMCell(n_hidden, forget_bias=1.0, state_is_tuple=True)
#            fw_cell = tf.contrib.rnn.GRUCell(n_hidden,activation=tf.tanh)
            lstm_fw_cell = tf.contrib.rnn.DropoutWrapper(fw_cell,output_keep_prob=dropout)
            lstm_fw_cell_m=tf.contrib.rnn.MultiRNNCell([lstm_fw_cell]*n_layers, state_is_tuple=True)
        # Backward direction cell
        with tf.name_scope("bwA"+scope),tf.variable_scope("bwA"+scope):
            print(tf.get_variable_scope().name)
            bw_cell = tf.contrib.rnn.LSTMCell(n_hidden, forget_bias=1.0, state_is_tuple=True)
#            bw_cell = tf.contrib.rnn.GRUCell(n_hidden, activation=tf.tanh)
            lstm_bw_cell = tf.contrib.rnn.DropoutWrapper(bw_cell,output_keep_prob=dropout)
            lstm_bw_cell_m = tf.contrib.rnn.MultiRNNCell([lstm_bw_cell]*n_layers, state_is_tuple=True)
        # Get lstm cell output
        #try:
        with tf.name_scope("bwA"+scope),tf.variable_scope("bwA"+scope):
            outputs, _, _ = tf.contrib.rnn.static_bidirectional_rnn(lstm_fw_cell_m, lstm_bw_cell_m, x, dtype=tf.float32)
            #         except Exception: # Old TensorFlow version only returns outputs not states
            #             outputs = tf.nn.bidirectional_rnn(lstm_fw_cell_m, lstm_bw_cell_m, x,
            #                                             dtype=tf.float32)
        # create a dense layer and get final output,noml and regulize to range [-1,1]
        output_temp = tf.layers.batch_normalization(tf.reduce_sum(outputs,0))
#         outputL2=tf.nn.l2_normalize(output_temp,1)
        with tf.name_scope("dwA"+scope),tf.variable_scope("denselayer_wA"+scope),tf.variable_scope("denselayer_bA"+scope),tf.variable_scope("denselayer_outputA"+scope):
            output = tf.nn.xw_plus_b(output_temp,
                             tf.get_variable("denselayer_wA", [2 * n_hidden, sequence_length]),
                             tf.get_variable("denselayer_bA", [sequence_length]))
            output = tf.nn.elu(output, "densepayer_outputA")
        output = tf.layers.batch_normalization(output)
#         with tf.name_scope("dw2"+scope),tf.variable_scope("denselayer_w2"+scope),tf.variable_scope("denselayer_b2"+scope),tf.variable_scope("denselayer_output2"+scope):
#             output2 = tf.nn.xw_plus_b(output,
#                              tf.get_variable("denselayer_w2", [n_hidden, 50]),
#                              tf.get_variable("denselayer_b2", [50]))
#             output2 = tf.nn.elu(output2, "densepayer_output2")
#         output2 = tf.layers.batch_normalization(output2)
        return output

    def contrastive_loss(self, y,d,batch_size):
        tmp= y *tf.square(d)
        #tmp= tf.mul(y,tf.square(d))
        tmp2 = (1-y) *tf.square(tf.maximum((1 - d),0))
        return tf.reduce_sum(tmp +tmp2)/batch_size/2

    def cosine_dis(self,out1,out2):
        distance = tf.sqrt(tf.reduce_sum(tf.square(tf.subtract(out1,out2)),1,keep_dims=True))
        distance = tf.div(distance, tf.add(tf.sqrt(tf.reduce_sum(tf.square(out1),1,keep_dims=True)),tf.sqrt(tf.reduce_sum(tf.square(out2),1,keep_dims=True))))
#         distance = tf.reshape(distance, [-1], name="distance")
        return distance

    def __init__(
      self, sequence_length, embedding_size, hidden_units, l2_reg_lambda, batch_size, word_embeddings = None):

      # Placeholders for input, output and dropout
      self.input_x1 = tf.placeholder(tf.int32, [None, sequence_length], name="input_x1")
      self.input_x2 = tf.placeholder(tf.int32, [None, sequence_length], name="input_x2")
      self.input_y = tf.placeholder(tf.int32, [None], name="input_y")
      self.dropout_keep_prob = tf.placeholder(tf.float32, name="dropout_keep_prob")
      num_class = 2

      # Keeping track of l2 regularization loss (optional)
      l2_loss = tf.constant(0.0, name="l2_loss")

      # Embedding layer
      with tf.name_scope("embedding"):
          if(word_embeddings == None):
              self.embedding = tf.Variable(tf.random_uniform([4200000, 100], -1.0, 1.0),trainable=True,name="W")
          else:
              rows = word_embeddings.shape[0]
              cols = word_embeddings.shape[1]
              self.embedding = tf.get_variable("embedding", [rows, cols], trainable=False)
          self.embedded_chars1 = tf.nn.embedding_lookup(self.embedding, self.input_x1)
          self.embedded_chars2 = tf.nn.embedding_lookup(self.embedding, self.input_x2)

     # Create a convolution + maxpool layer for each filter size
      with tf.name_scope("output"):
        self.out1_F=self.BiRNN_FeatureExtract(self.embedded_chars1, self.dropout_keep_prob, "side1", embedding_size, sequence_length)
        self.out1_A=self.BiRNN_AttentionMechanism(self.embedded_chars1, self.dropout_keep_prob, "side1", embedding_size, sequence_length)
        self.out2_F=self.BiRNN_FeatureExtract(self.embedded_chars2, self.dropout_keep_prob, "side2", embedding_size, sequence_length)
        self.out2_A=self.BiRNN_AttentionMechanism(self.embedded_chars2, self.dropout_keep_prob, "side2", embedding_size, sequence_length)
        self.cos_F = self.cosine_dis(self.out1_F,self.out2_F)
        self.cos_A = self.cosine_dis(self.out1_A,self.out2_A)

        self.out = tf.concat(axis=1, values=[self.out1_F,self.out1_A,self.out2_F, self.out2_A,self.cos_F,self.cos_A], name="finalOutput")
        logit = tf.nn.xw_plus_b(self.out,
                             tf.get_variable("dense_w1", [sequence_length * 4 + 2, sequence_length]),
                             tf.get_variable("dense_b1", [sequence_length]))
        logit = tf.nn.elu(logit, "dense_elu")
        logit = tf.layers.batch_normalization(logit)
        logits = tf.nn.xw_plus_b(logit,
                             tf.get_variable("dense_w2", [sequence_length, num_class]),
                             tf.get_variable("dense_b2", [num_class]))
        self.results = tf.argmax(logits, 1)

      with tf.name_scope("loss"):
          #calculate contrastive_loss, and the formula mentioned in paper: Learning Text Similarity with Siamese Recurrent Networks,but a similar difference: 1 / 4 to all
#         self.loss = self.contrastive_loss(self.input_y,self.results, batch_size)
        labels = tf.expand_dims(self.input_y, 1)
        indices = tf.expand_dims(tf.range(0, batch_size), 1)
        concated = tf.concat(axis=1,values=[indices, labels],name="one_hot_lable")
        onehot_labels = tf.sparse_to_dense(concated, tf.stack([batch_size, num_class]), 1.0, 0.0)
        loss_t = tf.nn.softmax_cross_entropy_with_logits(logits=logits,labels=onehot_labels,name='xentropy')
        self.loss = loss = tf.reduce_mean(loss_t, name='xentropy_mean')

#         self._lr = tf.Variable(0.0, trainable=False)
#         tvars = tf.trainable_variables()
#         grads, _ = tf.clip_by_global_norm(tf.gradients(cost, tvars),
#                                       config.max_grad_norm)
#         optimizer = tf.train.GradientDescentOptimizer(self.lr)
#         self._train_op = optimizer.apply_gradients(zip(grads, tvars))

      with tf.name_scope("accuracy"):
        correct_predictions = tf.equal(tf.cast(self.results, tf.int32), self.input_y)
        self.accuracy=tf.reduce_mean(tf.cast(correct_predictions, "float"), name="accuracy")