Towards Robust Online Dialogue Response Generation

Leyang Cui, Fandong Meng, Yijin Liu, Jie Zhou, Yue Zhang

Abstract:

Although pre-trained sequence-to-sequence models have achieved great success in dialogue response generation, chatbots still suffer from generating inconsistent responses in real-world applications, especially in multi-turn settings. We argue that this can be caused by a discrepancy between training and real-world testing. While the chatbot generates the response based on the gold context during training, it has to predict the next utterance based on the context consisting of both the user’s and the bot’s own utterances in real-world testing.
With the growing number of utterances, this discrepancy becomes more severe in the multi-turn settings. In this paper, we propose a hierarchical sampling-based method consisting of both utterance-level sampling and semi-utterance-level sampling, to alleviate the discrepancy, which increases the dialogue coherence implicitly. We further adopt reinforcement learning and re-ranking methods to explicitly optimize the dialogue coherence during training and inference, respectively. Empirical experiments show the effectiveness of the proposed methods for improving the robustness of chatbots in real practice.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Applications (eg, speech processing, computer vision, NLP)