the paper: Adversarial Learning for Neural Dialogue Generation https://arxiv.org/pdf/1701.06547.pdf

the paper translation in Chinese :http://blog.csdn.net/liuyuemaicha/article/details/60581187

TensorFlow 0.12.0 Python 2.7

gen_data: training data for gen model

disc_data: training data for disc model

disc: code about disc model

gen: code about gen model

utils: code about data operation and model config

notice:

gen_data include chitchat.train.answer, chitchat.train.query, chitchat.dev.answer, chitchat.dev.query (total four files)

disc_data include disc.dev.answer,disc.dev.query, disc.dev.gen 和 disc.train.answer, disc.train.query,disc.tran.gen (total six files)

formula of training data one sentence one row and splited with space, eg: i don ' t want to !

python al_neural_dialogue_train.py

introduction

def main(_):

'''

# step_1 training gen model

# gen_pre_train()
# model test
# gen_test()
# step_2 gen training data for disc
# gen_disc()
# step_3 training disc model
# disc_pre_train()
# step_4 training al model
# al_train()
# model test
# gen_test() 

'''

model introduction

1、disc model : hierarchical rnn (paper——Building end-to-end dialogue systems using generative hierarchical neural network models)

2、gen model : seq2seq model with attention (GRU cell)

3、method of reward : Monte Carlo Search

4、optimal:Policy Gradient