【发布时间】：2020-06-03 22:07:41
【问题描述】：
大家好，我为聊天机器人开发了两个 RNN 模型。假设用户说：“告诉我明天巴黎的天气如何”。 第一个模型将能够识别用户的意图 WEATHER_INFO ，而第二个模型将能够从 LOC:Paris 和 DATE:tomorrow 等短语中提取有意义的信息 当然还有许多其他意图类别，例如 MUSIC_PLAY 等等。 由于这两个模型没有以任何方式链接到同一个短语，我们可以得到例如结果 intent:WEATHER_INFO 和 TITLE:Paris ，其中插槽 TITLE 改为链接到 MUSIC_PLAY 意图 许多研究人员试图通过创建一个联合模型来提高性能，其中两个模型使用另一个模型以避免这种错误 这是我的插槽填充代码 nn
from keras.models import Model, Input
from keras.layers import LSTM, Embedding, Dense, TimeDistributed, Dropout, Bidirectional
import keras as k
from keras_contrib.layers import CRF
# input = Input(shape=(140,))
# input = Input(shape=(len(X_train),max_len))
input = Input(shape=(max_len,))
# word_embedding_size = 150
word_embedding_size = 150
n_words = len(token_ids)
# Embedding Layer
model = Embedding(input_dim=n_words, output_dim=word_embedding_size, input_length=max_len)(input)
# model = Embedding(input_dim=n_words, output_dim=word_embedding_size, input_length=140)(input)
# BI-LSTM Layer
model = Bidirectional(LSTM(units=word_embedding_size,
return_sequences=True,
dropout=0.5,
recurrent_dropout=0.5,
kernel_initializer=k.initializers.he_normal()))(model)
model = LSTM(units=word_embedding_size * 2,
return_sequences=True,
dropout=0.5,
recurrent_dropout=0.5,
kernel_initializer=k.initializers.he_normal())(model)
# TimeDistributed Layer
model = TimeDistributed(Dense(n_tags, activation="relu"))(model)
# CRF Layer
crf = CRF(n_tags)
out = crf(model) # output
model = Model(input, out)
# In[]
# model compile and fit
from keras.callbacks import ModelCheckpoint
import matplotlib.pyplot as plt
# Optimiser
adam = k.optimizers.Adam(lr=0.0005, beta_1=0.9, beta_2=0.999)
# Compile model
model.compile(optimizer=adam, loss=crf.loss_function, metrics=[crf.accuracy, 'accuracy'])
model.summary()
# Saving the best model only
filepath = "ner-bi-lstm-td-model-{val_accuracy:.2f}.hdf5"
checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True, mode='max')
callbacks_list = [checkpoint]
# Fit the best model
history = model.fit(X_train, y_train, batch_size=256, epochs=10, validation_split=0.1, verbose=1,
callbacks=callbacks_list)
这里是意图检测NN的代码
#CNN architecture
from __future__ import print_function
import keras
from keras.datasets import mnist
from keras.models import Sequential
from keras.layers import Dense, Dropout, Flatten
from keras.layers import Conv2D, MaxPooling2D
from keras import backend as K
from keras import layers
batch_size = 128
epochs = 12
if nn_architecture == 'CNN':
model_CNN = Sequential()
e = Embedding(vocab_size, 300, weights=[embedding_matrix], input_length=max_length, trainable=False)
model_CNN.add(e)
model_CNN.add(Dropout(0.2))
# we add a Convolution1D, which will learn filters
# word group filters of size filter_length:
filters = 50
kernel_size = 3
hidden_dims = 250
model_CNN.add(layers.Conv1D(filters,
kernel_size,
padding='valid',
activation='relu',
strides=1))
# we use max pooling:
model_CNN.add(layers.GlobalMaxPooling1D())
# We add a vanilla hidden layer:
model_CNN.add(Dense(hidden_dims))
model_CNN.add(Dropout(0.2))
model_CNN.add(layers.Activation('relu'))
# We project onto a single unit output layer, and squash it with a sigmoid:
model_CNN.add(Dense(nbClasses)) # no_of_categories
model_CNN.add(layers.Activation('sigmoid'))
model_CNN.compile(loss='binary_crossentropy',
optimizer='adam',
metrics=['accuracy'])
history_CNN =model_CNN.fit(X_train, Y_train_c,
batch_size=batch_size,
epochs=epochs,
#validation_split=0.2
)
# Epoch 12/12
# 38771/38771 [==============================] - 11s 276us/step -
#loss: 0.0046 - accuracy: 0.9985
请帮助我...提前谢谢
【问题讨论】：
标签： nlp recurrent-neural-network gated-recurrent-unit