Construction of a dialogue generation model based on persuasion strategies

Abstract

Persuasion is the act of inducing a change to audience member’s consciousnesses or behaviors. The dialogue from the persuader can be divided into several types depending on the approach, and the types are defined as persuasion strategies. The persuadee may react differently even with the same persuasion strategy of the persuader. It is necessary to respond differently depending on the reaction from the persuader for effective persuasion. More persuasive responses can be generated by reflecting the concept of persuasion strategy into the generation process of the model. The model that creates a more persuasive dialogue for a given discourse suggested in this work can interact with the persuader by reflecting the persuasion strategy to the response generation and classifying the persuasion strategy of the generated response. The generated response by the suggested model showed a number greater than the response by the baseline in each metric, and it was confirmed that the model could interact with the persuadee multiple times.