Topic Aware Context Modelling for Dialogue Response Generation

摘要

Response generation is an important direction in conversation systems. Currently a lot of approaches have been proposed and achieved significant improvement. However, an important limitation has been widely realized as most models tend to generate general answers. To cope with this limitation, besides the needs of more sophisticated generation models, how to use extra information is also an important direction. In this research, inspired by the importance of topics in conversation, we proposed a topic aware context modelling framework by utilizing similar question answer pairs in the repository. Furthermore, we use adversarial learning to improve the quality of generated response. The experimental study has shown the propose framework's potential.