One of the signs of listening attentively and supportively is occasional back-channel feedback, small utterances produced by the listener while the speaker continues his turn. To do this appropriately it is necessary to understand when back-channels are and are not welcome. In Egyptian Arabic, times when the listener is especially welcome to back-channel are indicated by various prosodic features produced by the speaker, including a steep pitch downslope. This particular feature contrasts with the downward pitch staircase (Kadenz) characteristic of turn-yields. This finding is based on qualitative and quantitative analysis of the contexts of occurrence of 660 back-channels in 168 minutes of Egyptian Arabic telephone dialogs from the Callhome corpus. Note: reflecting new developments, we have improved the paper slightly; moving section 8.2 into section 9, and rewriting and adding to sections 9.2 and 9.5. Corresponding Author: Nigel Ward nigelward@acm.org phone: 915-747-6827 fax: 915-747-5030 http://www.cs.utep.edu/nigel/ Computer Science, University of Texas at El Paso El Paso, TX 79968-0518 February 14, 2007 0Acknowledgements: We thank Thamar Solorio, W. Lewis Johnson, Jon Amastae and the participants of the 20th Arabic Linguistics Symposium for discussion. This work was supported in part by DARPA and in part by the National Science Foundation under Grant No. 0415150. A Prosodic Feature that Invites Back-Channels in Egyptian Arabic 1 Back-Channeling as a Dialog Skill To be a good listener you have to be able to show you’re listening. In dialog this includes the active display of attention, interest, understanding and/or willingness to let the other person continue. This is accomplished in part with back-channels, also known as “minimal responses” and “continuers”: the short utterances produced while the interlocutor has the turn. In English these are typically utterances such as uh-huh; in Egyptian Arabic the most common back-channels are ah, mmm, laughter, tayeb, and aiwa. This raises the question of how a listener can know when it is appropriate to produce a backchannel. Work in other languages (Yngve 1970; Ward & Tsukahara 2000; Fujie et al. 2005) suggests this depends on both speaker-related factors and listener-related factors. That is, the listener is free to produce back-channels based on his own understanding and intentions, but these back-channels are especially welcome at certain times in the dialog, and these times are determined by what the speaker is saying and how he is saying it. Further, these times are indicated in part by prosody: in several languages there is a prosodic cue that a speaker can use to indicate when he welcomes a back-channel from the listener. The present study was motivated by the desire to be able to teach Arabic back-channel skills to non-natives, specifically by extending an intelligent tutoring system, the Tactical Language Trainer (Johnson et al. 2005). The motivating problem is that a second language learner who lacks turn-taking skills, even if a master of the vocabulary and grammar, can easily appear uninterested, ill-informed, thoughtless, discourteous, passive, indecisive, untrusting, dull, pushy, or worse. Indeed, our earlier study of Japanese and English back-channel behavior showed that not only do Japanese back-channel twice as often as Americans (Maynard 1989), but that the interval between the prosodic cue from the speaker to the back-channel response by the listener was typically only half as long in Japanese (Ward & Tsukahara 2000). The potential for awkward intercultural interactions here is clear. Unfortunately the rules governing turn-taking are seldom taught to language learners, largely because they are not known. This has been the case for Arabic. This paper describes the initial identification of a prosodic cue in Egyptian Arabic that indicates to the interlocutor when back-channel feedback is especially welcome, and which makes it statistically more likely that the listener will indeed produce a back-channel in response. 2 Prosody and Turn-Taking in Arabic Research findings relating to our question are found in two areas: turn-taking and prosody. Back-channel behavior is an aspect of turn-taking, that is, the way that speakers in dialog manage their interactions to allow smooth exchanges and minimize awkward silences and interruptions. In Arabic, the only work on this is that of Hafez (1991), which provides a useful taxonomy of the ways in which speakers manage turn-taking in Egyptian Arabic. Hafez further identifies lexical discourse markers which often accompany turn taking and turn yielding, however unfortunately not for backchanneling. In his brief discussion of back-channels, Hafez provides examples of their semantic and pragmatic functions; these appear to be similar to those seen in other languages (Ward & Tsukahara 2000). In particular, some back-channels do not display understanding, but merely attention. Hafez also notes that back-channels can occur not only in “slots” (places where the interlocutor is momentarily silent) but also “in overlap” with the other speaker’s turn; again this is also seen in other languages.
[1]
Turn-taking in Egyptian Arabic: Spontaneous speech vs. drama dialogue
,
1991
.
[2]
Treebank Penn,et al.
Linguistic Data Consortium
,
1999
.
[3]
V. Yngve.
On getting a word in edgewise
,
1970
.
[4]
Stacy Marsella,et al.
Tactical Language Training System: An Interim Report
,
2004,
Intelligent Tutoring Systems.
[5]
Nigel G. Ward,et al.
A case study in the identification of prosodic cues to turn-taking: back-channeling in Arabic
,
2006,
INTERSPEECH.
[6]
Steven H. Lewis,et al.
Listener Responsiveness and the Coordination of Conversation
,
1982
.
[7]
Nigel G. Ward,et al.
Prosodic features which cue back-channel responses in English and Japanese
,
2000
.
[8]
W. Nigel,et al.
Pragmatic functions of prosodic features in non-lexical utterances
,
2004,
Speech Prosody 2004.
[9]
Iain R. Murray,et al.
Toward the simulation of emotion in synthetic speech: a review of the literature on human vocal emotion.
,
1993,
The Journal of the Acoustical Society of America.
[10]
A. Rajouani,et al.
ANALYSIS AND SYNTHESIS OF INTERROGATIVE INTONATION IN ARABIC
,
1999
.
[11]
David G. Novick,et al.
The UTEP Corpus of Iraqi Arabic
,
2006
.
[12]
Andreas Stolcke,et al.
Direct Modeling of Prosody: An Overview of Applications in Automatic Speech Processing
,
2004
.
[14]
Nigel Ward,et al.
Non-lexical conversational sounds in American English
,
2006
.
[15]
D. Bolinger.
Intonation and Its Parts
,
1985
.
[16]
Tetsunori Kobayashi,et al.
Back-channel feedback generation using linguistic and nonlinguistic information and its application to spoken dialogue system
,
2005,
INTERSPEECH.
[17]
Nigel Ward,et al.
Learning to Show You're Listening: A Back-Channel Trainer for Arabic
,
2007
.
[18]
D. Bolinger.
Intonation and Its Uses
,
1989
.
[19]
Khaled Rifaat,et al.
The Structure of Arabic Intonation: A preliminary investigation
,
2005
.
[20]
K. D. Jong,et al.
Stress, duration, and intonation in Arabic word-level prosody
,
1999
.