Update app.py

app.py

@@ -34,13 +34,14 @@ def model_(n_gram = 21):
     input1 = Input(shape=(21,),dtype='float32',name = 'char_input')
     input2 = Input(shape=(21,),dtype='float32',name = 'type_input')
 
-    a = Embedding(180, 32,input_length=21)(input1)
-    a = SpatialDropout1D(0.1)(a)
-    a = TimestepDropout(0.05)(a)
+    a = Embedding(178, 32,input_length=21)(input1)
+    a = SpatialDropout1D(0.15)(a)
+    #a = TimestepDropout(0.05)(a)
     char_input = BatchNormalization()(a)
 
     a_concat = []
-    filters = [[1,200],[2,200],[3,200],[4,200],[5,200],[6,200],[7,200],[8,200],[9,150],[10,150],[11,150],[12,100]]
+    filters = [[1,200],[2,200],[3,200],[4,200],[5,200],[6,200],[8,200],[11,150],[12,100]]
+    #filters = [[1,200],[2,200],[3,200],[4,200],[5,200],[6,200],[7,200],[8,200],[9,150],[10,150],[11,150],[12,100]]
     
     for (window_size, filters_size) in filters:
         convs = Conv1D(filters=filters_size, kernel_size=window_size, strides=1)(char_input)
@@ -49,19 +50,22 @@ def model_(n_gram = 21):
         convs = ZeroPadding1D(padding=(0, window_size-1))(convs)
         a_concat.append(convs)
     token_max = Maximum()(a_concat)
-    lstm_char = Bidirectional(LSTM(100 ,return_sequences=True))(char_input)
+    lstm_char = Bidirectional(LSTM(128 ,return_sequences=True,kernel_regularizer=regularizers.L2(0.0000001),bias_regularizer=regularizers.L2(0.0000001)))(char_input)
+    lstm_char = Dense(64, activation='elu')(lstm_char)
+    #lstm_char = Bidirectional(LSTM(64 ,return_sequences=True))(lstm_char)
+    #lstm_char = Attention(return_sequences=True)(lstm_char)
     
     b = Embedding(12, 12, input_length=21)(input2)
-    b = SpatialDropout1D(0.1)(b)
-    type_inputs = TimestepDropout(0.05)(b)
+    type_inputs = SpatialDropout1D(0.15)(b)
+    #type_inputs = TimestepDropout(0.05)(b)
 
-    x = Concatenate()([lstm_char, type_inputs, char_input, token_max])
+    x = Concatenate()([type_inputs, char_input, lstm_char, token_max])
     x = BatchNormalization()(x)
-
+    
     x = Flatten()(x)
-    x = Dense(200, activation='elu')(x)
+    x = Dense(100, activation='elu')(x)
     x = Dropout(0.2)(x)
-    out = Dense(1, activation='sigmoid',dtype = 'float32')(x)
+    out = Dense(1, activation='sigmoid',dtype = 'float32',kernel_regularizer=regularizers.L2(0.01),bias_regularizer=regularizers.L2(0.01))(x)
     
 
     model = Model(inputs=[input1, input2], outputs=out)
@@ -101,7 +105,7 @@ def tokenize(text):
         word_end = []
  
         y_predict = model.predict([x_char, x_type], batch_size = 512)
-        y_predict = (y_predict.ravel() > 0.4609375).astype(int)
+        y_predict = (y_predict.ravel() > 0.46542968749999997).astype(int)
         word_end = y_predict[1:].tolist() + [1]
         
         tokens = []