fix syn_words and CRF

Author: yongzhuo

AugmentText/augment_eda/enhance_eda_v2.py

@@ -12,7 +12,7 @@
 
 
 KEY_WORDS = ["macropodus"] # 不替换同义词的词语
-ENGLISH = 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ'
+ENGLISH = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ"
 
 
 def is_english(text):
@@ -22,7 +22,7 @@ def is_english(text):
     :return: boolean, True or False
     """
     try:
-        text_r = text.replace(' ', '').strip()
+        text_r = text.replace(" ", "").strip()
         for tr in text_r:
             if tr in ENGLISH:
                 continue
@@ -39,7 +39,7 @@ def is_number(text):
     :return: boolean, True or False 
     """
     try:
-        text_r = text.replace(' ', '').strip()
+        text_r = text.replace(" ", "").strip()
         for tr in text_r:
             if tr.isdigit():
                 continue
@@ -57,7 +57,7 @@ def get_syn_word(word):
     """
     if not is_number(word.strip()) or not is_english(word.strip()):
         word_syn = synonyms.nearby(word)
-        word_syn = word_syn if not word_syn else [word]
+        word_syn = word_syn[0] if len(word_syn[0]) else [word]
         return word_syn
     else:
         return [word]
@@ -124,7 +124,7 @@ def word_swap(words, n=1):
     while count < n:
         idx_select = random.sample(idxs, 2)
         temp = words[idx_select[0]]
-        words[idx_select[0]] =  words[idx_select[1]]
+        words[idx_select[0]] = words[idx_select[1]]
         words[idx_select[1]] = temp
         count += 1
     return words
@@ -182,7 +182,7 @@ def eda(text, n=1, use_syn=True):
     return sens_4
 
 
-if __name__ == '__main__':
+if __name__ == "__main__":
     sens = "".join(["macropodus", "是不是", "哪个", "啦啦",
                     "只需做好这四点，就能让你养的天竺葵全年花开不断！"])
     print(eda(sens))

Ner/bert/keras_bert_layer.py

@@ -23,7 +23,7 @@
 from keras import activations
 from keras import constraints
 import warnings
-import keras
+import os
 # crf_loss
 from keras.losses import sparse_categorical_crossentropy
 from keras.losses import categorical_crossentropy
@@ -220,7 +220,7 @@ def to_tuple(shape):
     with keras-team/keras. So we must apply this function to
     all input_shapes of the build methods in custom layers.
     """
-    if is_tf_keras:
+    if os.environ.get("TF_KERAS")==1:
         import tensorflow as tf
         return tuple(tf.TensorShape(shape).as_list())
     else:
@@ -530,36 +530,36 @@ def get_config(self):
         base_config = super(CRF, self).get_config()
         return dict(list(base_config.items()) + list(config.items()))
 
-    # @property
-    # def loss_function(self):
-    #     warnings.warn('CRF.loss_function is deprecated '
-    #                   'and it might be removed in the future. Please '
-    #                   'use losses.crf_loss instead.')
-    #     return crf_loss
-    #
-    # @property
-    # def accuracy(self):
-    #     warnings.warn('CRF.accuracy is deprecated and it '
-    #                   'might be removed in the future. Please '
-    #                   'use metrics.crf_accuracy')
-    #     if self.test_mode == 'viterbi':
-    #         return crf_viterbi_accuracy
-    #     else:
-    #         return crf_marginal_accuracy
-    #
-    # @property
-    # def viterbi_acc(self):
-    #     warnings.warn('CRF.viterbi_acc is deprecated and it might '
-    #                   'be removed in the future. Please '
-    #                   'use metrics.viterbi_acc instead.')
-    #     return crf_viterbi_accuracy
-    #
-    # @property
-    # def marginal_acc(self):
-    #     warnings.warn('CRF.moarginal_acc is deprecated and it '
-    #                   'might be removed in the future. Please '
-    #                   'use metrics.marginal_acc instead.')
-    #     return crf_marginal_accuracy
+    @property
+    def loss_function(self):
+        warnings.warn('CRF.loss_function is deprecated '
+                      'and it might be removed in the future. Please '
+                      'use losses.crf_loss instead.')
+        return crf_loss
+
+    @property
+    def accuracy(self):
+        warnings.warn('CRF.accuracy is deprecated and it '
+                      'might be removed in the future. Please '
+                      'use metrics.crf_accuracy')
+        if self.test_mode == 'viterbi':
+            return crf_viterbi_accuracy
+        else:
+            return crf_marginal_accuracy
+
+    @property
+    def viterbi_acc(self):
+        warnings.warn('CRF.viterbi_acc is deprecated and it might '
+                      'be removed in the future. Please '
+                      'use metrics.viterbi_acc instead.')
+        return crf_viterbi_accuracy
+
+    @property
+    def marginal_acc(self):
+        warnings.warn('CRF.moarginal_acc is deprecated and it '
+                      'might be removed in the future. Please '
+                      'use metrics.marginal_acc instead.')
+        return crf_marginal_accuracy
 
     @staticmethod
     def softmaxNd(x, axis=-1):
@@ -655,9 +655,22 @@ def step(self, input_energy_t, states, return_logZ=True):
             chain_energy = chain_energy * K.expand_dims(
                 K.expand_dims(m[:, 0] * m[:, 1]))
         if return_logZ:
-            # shapes: (1, B, F) + (B, F, 1) -> (B, F, F)
+            # # shapes: (1, B, F) + (B, F, 1) -> (B, F, F)
+            # energy = chain_energy + K.expand_dims(input_energy_t - prev_target_val, 2)
+            # new_target_val = K.logsumexp(-energy, 1)  # shapes: (B, F)
+            # return new_target_val, [new_target_val, i + 1]
+
             energy = chain_energy + K.expand_dims(input_energy_t - prev_target_val, 2)
-            new_target_val = K.logsumexp(-energy, 1)  # shapes: (B, F)
+            new_target_val = K.logsumexp(-energy, 1)
+            # added from here
+            if len(states) > 3:
+                if K.backend() == 'theano':
+                    m = states[3][:, t:(t + 2)]
+                else:
+                    m = K.slice(states[3], [0, t], [-1, 2])
+                is_valid = K.expand_dims(m[:, 0])
+                new_target_val = is_valid * new_target_val + (1 - is_valid) * prev_target_val
+            # added until here
             return new_target_val, [new_target_val, i + 1]
         else:
             energy = chain_energy + K.expand_dims(input_energy_t + prev_target_val, 2)