StyleDGPT: Stylized Response Generation with Pre-trained Language Models
This work addresses the challenge of creating more versatile dialogue systems for applications requiring specific stylistic outputs, though it is incremental as it builds on existing pre-trained models.
The paper tackled the problem of generating stylized responses in open-domain dialogue systems without parallel training data by fine-tuning pre-trained language models with a KL loss and style classifier. The result was a model that significantly outperformed state-of-the-art methods in style consistency and contextual coherence on two public datasets.
Generating responses following a desired style has great potentials to extend applications of open-domain dialogue systems, yet is refrained by lacking of parallel data for training. In this work, we explore the challenging task with pre-trained language models that have brought breakthrough to various natural language tasks. To this end, we introduce a KL loss and a style classifier to the fine-tuning step in order to steer response generation towards the target style in both a word-level and a sentence-level. Comprehensive empirical studies with two public datasets indicate that our model can significantly outperform state-of-the-art methods in terms of both style consistency and contextual coherence.