Abstract | ||
---|---|---|
This paper presents a Discriminative Deep Dyna-Q (D3Q) approach to improving the effectiveness and robustness of Deep Dyna-Q (DDQ), a recently proposed framework that extends the Dyna-Q algorithm to integrate planning for task-completion dialogue policy learning. To obviate DDQu0027s high dependency on the quality of simulated experiences, we incorporate an RNN-based discriminator in D3Q to differentiate simulated experience from real user experience in order to control the quality of training data. Experiments show that D3Q significantly outperforms DDQ by controlling the quality of simulated experience used for planning. The effectiveness and robustness of D3Q is further demonstrated in a domain extension setting, where the agentu0027s capability of adapting to a changing environment is tested. |
Year | Venue | DocType |
---|---|---|
2018 | EMNLP | Conference |
Volume | Citations | PageRank |
abs/1808.09442 | 2 | 0.35 |
References | Authors | |
14 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Shang-Yu Su | 1 | 9 | 4.88 |
Xiujun Li | 2 | 139 | 11.73 |
Jianfeng Gao | 3 | 5729 | 296.43 |
Jingjing Liu | 4 | 515 | 39.31 |
Yun-Nung Chen | 5 | 324 | 35.41 |