from nlp_toolkit.models import Base_Model
from nlp_toolkit.modules.logits import tc_output_logits
from nlp_toolkit.modules.token_embedders import Token_Embedding
from keras.layers import Input, Dense, Flatten, Dropout
from keras.layers import Conv1D, MaxPooling1D
from keras.layers.merge import concatenate
from keras.models import Model
class textCNN(Base_Model):
"""
The known Kim CNN model used in text classification.
It use mulit-channel CNN to encode texts
"""
def __init__(self, nb_classes, nb_tokens, maxlen,
embedding_dim=256, embeddings=None, embed_l2=1E-6,
conv_kernel_size=[3, 4, 5], pool_size=[2, 2, 2],
nb_filters=128, fc_size=128,
embed_dropout_rate=0.25, final_dropout_rate=0.5):
super(textCNN).__init__()
self.nb_classes = nb_classes
self.nb_tokens = nb_tokens
self.maxlen = maxlen
self.embedding_dim = embedding_dim
self.nb_filters = nb_filters
self.pool_size = pool_size
self.conv_kernel_size = conv_kernel_size
self.fc_size = fc_size
self.final_dropout_rate = final_dropout_rate
self.embed_dropout_rate = embed_dropout_rate
# core layer: multi-channel cnn-pool layers
self.cnn_list = [Conv1D(
nb_filters, f, padding='same', name='conv_%d' % k) for k, f in enumerate(conv_kernel_size)]
self.pool_list = [MaxPooling1D(p, name='pool_%d' % k)
for k, p in enumerate(pool_size)]
self.fc = Dense(fc_size, activation='relu',
kernel_initializer='he_normal')
if embeddings is not None:
self.token_embeddings = [embeddings]
else:
self.token_embeddings = None
self.invalid_params = {'cnn_list', 'pool_list', 'fc'}
def forward(self):
model_input = Input(shape=(self.maxlen,), dtype='int32', name='token')
x = Token_Embedding(model_input, self.nb_tokens, self.embedding_dim,
self.token_embeddings, False, self.maxlen,
self.embed_dropout_rate, name='token_embeddings')
cnn_combine = []
for i in range(len(self.conv_kernel_size)):
cnn = self.cnn_list[i](x)
pool = self.pool_list[i](cnn)
cnn_combine.append(pool)
x = concatenate(cnn_combine, axis=-1)
x = Flatten()(x)
x = Dropout(self.final_dropout_rate)(x)
x = self.fc(x)
outputs = tc_output_logits(x, self.nb_classes, self.final_dropout_rate)
self.model = Model(inputs=model_input,
outputs=outputs, name="TextCNN")
def get_loss(self):
if self.nb_classes == 2:
return 'binary_crossentropy'
elif self.nb_classes > 2:
return 'categorical_crossentropy'
def get_metrics(self):
return ['acc']
