UniGDD: A Unified Generative Framework for Goal-Oriented Document-Grounded Dialogue
This addresses a specific problem in dialogue systems for applications requiring document-based responses, but it is incremental as it builds on existing sub-task decomposition approaches.
The paper tackles goal-oriented document-grounded dialogue by proposing a unified generative framework to avoid error propagation from pipeline methods, achieving improved performance as demonstrated in experiments.
The goal-oriented document-grounded dialogue aims at responding to the user query based on the dialogue context and supporting document. Existing studies tackle this problem by decomposing it into two sub-tasks: knowledge identification and response generation. However, such pipeline methods would unavoidably suffer from the error propagation issue. This paper proposes to unify these two sub-tasks via sequentially generating the grounding knowledge and the response. We further develop a prompt-connected multi-task learning strategy to model the characteristics and connections of different tasks and introduce linear temperature scheduling to reduce the negative effect of irrelevant document information. Experimental results demonstrate the effectiveness of our framework.