Demo: Have a Chat with Sensitive Artificial Listeners

Abstract. Sensitive Artificial Listeners (SAL) are virtual dialoguepartners based on audiovisual analysis and synthesis. Despite theirvery limited verbal understanding, they intend to engage the user ina conversation by paying attention to the user’s emotions and non-verbal expressions. The SAL characters have their own emotionallydefined personality, and attempt to drag the user towards their dom-inant emotion through a combination of verbal and non-verbal ex-pression. The demonstrator shows the publicly available, fully au-tonomous SAL system. 1 INTRODUCTION The SAL demo supports sustained emotionally-coloured machine-human interaction using non-verbal expression. The system aims toengage the user in a dialogue by paying attention to the user’s non-verbal expression, and reacting accordingly. It focuses on the “softskills” that humans naturally use to keep a conversation alive.To simplify the challenge somewhat, the SAL system avoids task-oriented dialogue. Instead, it models the type of interaction foundat parties: you listen to someone you want to chat with, and withoutreally understanding much of what they are saying, you exhibit all thesigns that are needed for them to continue talking to you. Similarlyto the Rapport agent [2], SAL characters show non-verbal listenersignals; in addition, they can also speak to engage the user in a simpledialogue. The approach has been test-run using Wizard of Oz setupsat various stages of maturity [1]. This has allowed us to fine-tunethe scripts used by the various characters, in order to react to theemotional state of the user in plausible ways despite the lack of verbalknowledge. The system is freely available from the project websitewww.semaine-project.eu.