PLATO: Pre-trained Dialogue Generation Model with Discrete Latent Variable
This work addresses the one-to-many mapping challenge in dialogue generation for applications like chit-chat and question answering, though it is incremental as it builds on existing pre-training methods.
The authors tackled the problem of generating diverse and appropriate responses in dialogue systems by proposing PLATO, a pre-trained model with discrete latent variables, which achieved state-of-the-art results on multiple datasets.
Pre-training models have been proved effective for a wide range of natural language processing tasks. Inspired by this, we propose a novel dialogue generation pre-training framework to support various kinds of conversations, including chit-chat, knowledge grounded dialogues, and conversational question answering. In this framework, we adopt flexible attention mechanisms to fully leverage the bi-directional context and the uni-directional characteristic of language generation. We also introduce discrete latent variables to tackle the inherent one-to-many mapping problem in response generation. Two reciprocal tasks of response generation and latent act recognition are designed and carried out simultaneously within a shared network. Comprehensive experiments on three publicly available datasets verify the effectiveness and superiority of the proposed framework.