Learning to Predict Persona Information forDialogue Personalization without Explicit Persona Description
This addresses the limitation of reliance on explicit persona descriptions in dialogue systems, enabling broader application, though it is incremental as it builds on existing personalization methods.
The paper tackles the problem of personalizing dialogue agents without needing explicit persona descriptions during inference, by learning to predict persona information from dialogue history. Experimental results on the PersonaChat dataset show improvements in response consistency when using predicted self-persona and in engagingness when using predicted partner persona, with successful transfer to other datasets.
Personalizing dialogue agents is important for dialogue systems to generate more specific, consistent, and engaging responses. However, most current dialogue personalization approaches rely on explicit persona descriptions during inference, which severely restricts its application. In this paper, we propose a novel approach that learns to predict persona information based on the dialogue history to personalize the dialogue agent without relying on any explicit persona descriptions during inference. Experimental results on the PersonaChat dataset show that the proposed method can improve the consistency of generated responses when conditioning on the predicted profile of the dialogue agent (i.e. "self persona"), and improve the engagingness of the generated responses when conditioning on the predicted persona of the dialogue partner (i.e. "their persona"). We also find that a trained persona prediction model can be successfully transferred to other datasets and help generate more relevant responses.