Reinforcement learning (RL) is an effective approach to learn an optimal dialog policy for task-oriented visual dialog systems. A common practice is to apply RL on a neural sequence-to-sequence (seq2seq) framework with the action space being the output vocabulary in the decoder. However, it is difficult to design a reward function that can achieve a balance between learning an effective policy and generating a natural dialog response. This paper proposes a novel framework that alternatively trains a RL policy for image guessing and a supervised seq2seq model to improve dialog generation quality. We evaluate our framework on the GuessWhich task and the framework achieves the state-of-the-art performance in both task completion and dialog quality.
@article{arxiv.1909.05365,
title = {Building Task-Oriented Visual Dialog Systems Through Alternative Optimization Between Dialog Policy and Language Generation},
author = {Mingyang Zhou and Josh Arnold and Zhou Yu},
journal= {arXiv preprint arXiv:1909.05365},
year = {2019}
}
Comments
updated with acknowledgement and minor typo fixes on tables