Back-channel feedback generation using linguistic and nonlinguistic information and its application to spoken dialogue system

Challenges in Building Highly Interactive Dialogue Systems

“…Consider for example back‐channeling behavior. This is a prototypical interactive behavior, probably the best studied one, and one already implemented in several systems that back channel at appropriate times in response to user speech (Ward and Tsukahara 1999; Fujie, Fukushima, and Kobayashi 2005; Gratch et al 2007; Schröder et al 2012; Kawahara et al 2015). Back‐channeling demonstrations today work if the user has been guided to perform a specific type of dialogue, such as retelling a story, solving a puzzle, or engaging in small talk.…”

Nonverbal Cues in Human–Robot Interaction: A Communication Studies Perspective

“…Non-linguistic utterances and other forms of vocal but non-speech paralanguage can get around the sticky issue of natural language processing (NLP). Another promising direction is robotic regulators for backchannelling, or verbal interjections, and its various forms [113]. Backchannelling can be very frequent, as in aizuchi (interjections to indicate that the listener is paying attention) during Japanese conversations, to less frequent, as well as being comprised of a mix of verbal and nonverbal cues.…”

Dialogue

“…For clues that help in predicting TRPs, see amongst others Thórisson (2002), Schlangen (2006), Atterer et al (2008), Raux and Eskenazi (2009), and Gravano and Hirschberg (2011). Traum and Rickel (2002), Kronlid (2006), Selfridge and Heeman (2010), and Bohus and Horvitz (2011) describe computational models of turn taking in multi-party dialogue, while Cathcart et al (2003), Fujie et al (2005), and Meena et al (2014) offer models of backchannel placing.…”