Retrieval-Generation Alignment for End-to-End Task-Oriented Dialogue System
This work addresses the challenge of enhancing response generation in task-oriented dialogue systems, which is incremental as it builds on existing generative models like T5 and ChatGPT by refining retrieval and generation alignment.
The paper tackles the problem of generative models struggling to differentiate subtle differences among retrieved knowledge base records in task-oriented dialogue systems, resulting in improved response quality by using maximal marginal likelihood to train a perceptive retriever and incorporating meta knowledge to guide the generator.
Developing an efficient retriever to retrieve knowledge from a large-scale knowledge base (KB) is critical for task-oriented dialogue systems to effectively handle localized and specialized tasks. However, widely used generative models such as T5 and ChatGPT often struggle to differentiate subtle differences among the retrieved KB records when generating responses, resulting in suboptimal quality of generated responses. In this paper, we propose the application of maximal marginal likelihood to train a perceptive retriever by utilizing signals from response generation for supervision. In addition, our approach goes beyond considering solely retrieved entities and incorporates various meta knowledge to guide the generator, thus improving the utilization of knowledge. We evaluate our approach on three task-oriented dialogue datasets using T5 and ChatGPT as the backbone models. The results demonstrate that when combined with meta knowledge, the response generator can effectively leverage high-quality knowledge records from the retriever and enhance the quality of generated responses. The codes and models of this paper are available at https://github.com/shenwzh3/MK-TOD.