Persona Authentication through Generative Dialogue

TL;DR

This paper introduces a novel method for persona authentication in dialogue systems, using a learned conversational policy that verifies persona consistency through adaptive questioning, with proven mutual information maximization.

Contribution

It proposes a new learning objective for persona authentication and develops an adaptive questioning method that effectively verifies unseen persona profiles.

Findings

Effective question sequences discovered for persona verification

Method generalizes well to unseen profiles

Maximizes mutual information between dialogue and persona

Abstract

In this paper we define and investigate the problem of \emph{persona authentication}: learning a conversational policy to verify the consistency of persona models. We propose a learning objective and prove (under some mild assumptions) that local density estimators trained under this objective maximize the mutual information between persona information and dialog trajectory. Based on the proposed objective, we develop a method of learning an authentication model that adaptively outputs personalized questions to reveal the underlying persona of its partner throughout the course of multi-turn conversation. Experiments show that our authentication method discovers effective question sequences that generalize to unseen persona profiles.