Toward Stance-based Personas for Opinionated Dialogues

TL;DR

This paper introduces a novel stance-based persona representation for opinionated dialogues, aiming to capture deeper human traits like opinions and beliefs to improve language generation coherence.

Contribution

It proposes a new dataset and approach for stance-based personas, advancing beyond fact-based profiles to include complex personality aspects.

Findings

Stance-based personas better capture abstract personality traits.

The new dataset enables exploration of profound persona representations.

Improved claim generation quality with stance-based profiles.

Abstract

In the context of chit-chat dialogues it has been shown that endowing systems with a persona profile is important to produce more coherent and meaningful conversations. Still, the representation of such personas has thus far been limited to a fact-based representation (e.g. "I have two cats."). We argue that these representations remain superficial w.r.t. the complexity of human personality. In this work, we propose to make a step forward and investigate stance-based persona, trying to grasp more profound characteristics, such as opinions, values, and beliefs to drive language generation. To this end, we introduce a novel dataset allowing to explore different stance-based persona representations and their impact on claim generation, showing that they are able to grasp abstract and profound aspects of the author persona.