A shallow model of backchannel continuers in spoken dialogue

Abstract

Spoken dialogue systems would be more acceptable if they were able to produce backchannel continuers such as mm-hmm in naturalistic locations during the user's utterances. Using the HCRC Map Task Corpus as our data source, we describe models for predicting these locations using only limited processing and features of the user's speech that are commonly available, and which therefore could be used as a low-cost improvement for current systems. The baseline model inserts continuers after a predetermined number of words. One further model correlates back-channel continuers with pause duration, while a second predicts their occurrence using trigram POS frequencies. Combining these two models gives the best results.