A model of back-channel acknowledgements in spoken dialogue

A Shallow Model of Backchannel Continuers in Spoken Dialogue Nicola Cathcart� Canon Research Centre Europe�Jean Carletta and Ewan Klein School of Informatics Bracknell, UK� University of Edinburgh nicolac@cre.canon.co.uk� {jeanc,ewan}@inf.ed.ac.uk 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 utter- ances. Using the HCRC Map Task Cor- pus as our data source, we describe mod- els 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 pre- determined number of words. One fur- ther model correlates back-channel contin- uers with pause duration, while a second pre- dicts their occurrence using trigram POS fre- quencies. Combining these two models gives the best results. 1 Introduction In a spoken dialogue between people, the participants use simple utterances such as yeah, a totty wee bit aye and mm-hnint to signal that communication is work- ing. Without this feedback, the partner may assume that he has not been understood and reformulate his ut- terance. Following Yngve (1970), we will use the term backchannel for such utterances. Although these can be substantive because they can repeat material from the partner's utterance (Clark and Schaefer, 1991), e.g., Right, okay, I'm below the fiat rocks, we will adopt (Jo- rafsky et al., 1998)'s terminology of continuer. We will take this to refer to the class of backchannel ut- terances, with minimal content, used to clearly signal that the speaker should continue with her current turn. (Yankelovich et al., 1995) point out that users of speech interface systems need feedback, too, especially since the system's silence could mean either of two very dif- ferent things: that it is waiting for user input, in which case the user should speak, or that it is still processing information, in which case the user should not. How- ever, any feedback must come at the right time or else it risks disrupting the speaker and ultimately, delaying task completion (Hirasawa et al., 1999). Most of our data, including the examples given above, are drawn from the HCRC Map Task Corpus, described in more detail in Section 3. Clearly these di- alogues are significantly more complex than the kind of interactions supported by current commercial spoken dialogue systems, where the length of user utterances is severely constrained. What kind of system would in- volve potentially lengthy user instructions comparable to those found in the Map Task? Lauria et al. (2001), Lemon et al. (2002), and Theobalt et al. (2002) describe work on building spoken dialogue systems for convers- ing with mobile robots, and this is a setting where com- plex instructions naturally arise. For example, in one scenario, 1 users attempt to teach routes and route seg- ments to a robot. (1) is a portion of such an instruction. (1) okay go to the end of the road and …

Content truncated to 500 words