Boosting Naturalness of Language in Task-oriented Dialogues via Adversarial Training
This work addresses the need for more human-like responses in task-oriented dialogue systems, which is incremental as it builds on existing adversarial training methods.
The paper tackled the problem of generating natural and fluent language in task-oriented dialogues by integrating adversarial training, resulting in a model that outperformed the previous state-of-the-art by 3.6% in BLEU on the RNN-LG Restaurant dataset.
The natural language generation (NLG) module in a task-oriented dialogue system produces user-facing utterances conveying required information. Thus, it is critical for the generated response to be natural and fluent. We propose to integrate adversarial training to produce more human-like responses. The model uses Straight-Through Gumbel-Softmax estimator for gradient computation. We also propose a two-stage training scheme to boost performance. Empirical results show that the adversarial training can effectively improve the quality of language generation in both automatic and human evaluations. For example, in the RNN-LG Restaurant dataset, our model AdvNLG outperforms the previous state-of-the-art result by 3.6% in BLEU.