def RNN_Model(max_length,
num_classes,
num_features,
embedding_matrix=None,
trainable=False,
input_shape=None,
input_tensor=None,
filter_num=64,
emb_size=300):
in_text = Input(shape=(max_length, ))
op_units, op_activation = _get_last_layer_units_and_activation(
num_classes, use_softmax=True)
trainable = True
if embedding_matrix is None:
x = Embedding(num_features, 64, trainable=trainable)(in_text)
else:
x = Embedding(num_features,
300,
trainable=trainable,
weights=[embedding_matrix])(in_text)
x = CuDNNGRU(128, return_sequences=True)(x)
x = GlobalMaxPooling1D()(x)
x = Dense(128)(x) #
x = PReLU()(x)
x = Dropout(0.35)(x) # 0
x = BatchNormalization()(x)
y = Dense(op_units, activation=op_activation)(x)
md = keras.models.Model(inputs=[in_text], outputs=y)
return md
def TextCNN_Model(input_shape,
embedding_matrix,
max_length,
num_features,
num_classes,
input_tensor=None,
filter_num=64,
emb_size=300,
trainable=False,
use_multi_label=False):
op_units, op_activation = _get_last_layer_units_and_activation(
num_classes, use_softmax=True)
in_text = Input(name='inputs', shape=[max_length], tensor=input_tensor)
if embedding_matrix is None:
layer = Embedding(input_dim=num_features,
output_dim=emb_size,
input_length=input_shape)(in_text)
else:
layer = Embedding(input_dim=num_features,
output_dim=emb_size,
input_length=input_shape,
weights=[embedding_matrix],
trainable=trainable)(in_text)
cnns = []
filter_sizes = [2, 3, 4, 5]
for size in filter_sizes:
cnn_l = Conv1D(filter_num,
size,
padding='same',
strides=1,
activation='relu')(layer)
pooling_0 = MaxPooling1D(max_length - size + 1)(cnn_l)
pooling_0 = Flatten()(pooling_0)
cnns.append(pooling_0)
cnn_merge = concatenate(cnns, axis=-1)
out = Dropout(0.2)(cnn_merge)
if use_multi_label:
main_output = Dense(op_units, activation='sigmoid')(out)
else:
main_output = Dense(op_units, activation=op_activation)(out)
md = keras.models.Model(inputs=in_text, outputs=main_output)
return md
def TextRCNN_Model(input_shape,
embedding_matrix,
max_length,
num_features,
num_classes,
input_tensor=None,
emb_size=300,
filter_num=64,
rnn_units=128,
trainable=False):
inputs = Input(name='inputs', shape=[max_length], tensor=input_tensor)
op_units, op_activation = _get_last_layer_units_and_activation(
num_classes, use_softmax=True)
if embedding_matrix is None:
layer = Embedding(input_dim=num_features,
output_dim=emb_size,
input_length=input_shape)(inputs)
else:
layer = Embedding(input_dim=num_features,
output_dim=emb_size,
weights=[embedding_matrix],
trainable=trainable)(inputs)
layer_cell = CuDNNGRU
embedding_output = layer
# 拼接
x_feb = Bidirectional(layer_cell(units=rnn_units,
return_sequences=True))(embedding_output)
x_feb = Concatenate(axis=2)([x_feb, embedding_output])
####使用多个卷积核##################################################
x_feb = Dropout(rate=0.5)(x_feb)
dim_2 = K.int_shape(x_feb)[2]
len_max = max_length
x_feb_reshape = Reshape((len_max, dim_2, 1))(x_feb)
# 提取n-gram特征和最大池化， 一般不用平均池化
conv_pools = []
filters = [2, 3, 4, 5]
for filter_size in filters:
conv = Conv2D(
filters=filter_num,
kernel_size=(filter_size, dim_2),
padding='valid',
kernel_initializer='normal',
activation='relu',
)(x_feb_reshape)
print("check conv", conv.get_shape())
pooled = MaxPooling2D(
pool_size=(len_max - filter_size + 1, 1),
strides=(1, 1),
padding='valid',
)(conv)
print("check pooled", pooled.get_shape())
conv_pools.append(pooled)
# 拼接
x = Concatenate()(conv_pools)
x = Flatten()(x)
#########################################################################
output = Dense(op_units, activation=op_activation)(x)
md = keras.models.Model(inputs=inputs, outputs=output)
return md