0 évaluation0% ont trouvé ce document utile (0 vote)
9 vues1 page
1. The study investigated the use of the response token "uhu" in naturally occurring conversations to see if prosodic cues like pitch patterns differentiated between instances where it functioned as a continuer versus an acknowledgment.
2. No differences were found in pitch range or movement between the two functions. Pitch characteristics varied within classes but did not predict the interactional differences.
3. Pitch characteristics of a response token appeared to depend on the pitch of the prior speaker's talk, copying a rise or fall. This copying behavior seemed to manage turn-taking in conversations.
4. A technique using cross-correlation to objectively measure prosodic similarity between pitch contours is presented and could be extended to other cues
1. The study investigated the use of the response token "uhu" in naturally occurring conversations to see if prosodic cues like pitch patterns differentiated between instances where it functioned as a continuer versus an acknowledgment.
2. No differences were found in pitch range or movement between the two functions. Pitch characteristics varied within classes but did not predict the interactional differences.
3. Pitch characteristics of a response token appeared to depend on the pitch of the prior speaker's talk, copying a rise or fall. This copying behavior seemed to manage turn-taking in conversations.
4. A technique using cross-correlation to objectively measure prosodic similarity between pitch contours is presented and could be extended to other cues
1. The study investigated the use of the response token "uhu" in naturally occurring conversations to see if prosodic cues like pitch patterns differentiated between instances where it functioned as a continuer versus an acknowledgment.
2. No differences were found in pitch range or movement between the two functions. Pitch characteristics varied within classes but did not predict the interactional differences.
3. Pitch characteristics of a response token appeared to depend on the pitch of the prior speaker's talk, copying a rise or fall. This copying behavior seemed to manage turn-taking in conversations.
4. A technique using cross-correlation to objectively measure prosodic similarity between pitch contours is presented and could be extended to other cues
Prosodic matching of response tokens in conversational speech
Jan Gorisch, University of Sheffield,
Department of Computer Science, Department of Human Communication Sciences The phonetic forms and discourse functions of response tokens such as mm have been investigated for several decades !n general, previous studies have attempted to classify response tokens according to their communicative function, and then to search for prosodic similarities "ithin each class #Gardner, $%%&' Using data from naturally(occurring research meetings in the )*! corpus #native +ritish and )merican speakers' "e investigated the response token uhu in this "ay !nteractional analysis "as used to identify instances of uhu that functioned either as continuers or as ackno"ledgments, and then "e attempted to identify prosodic #and visual' cues "hich might differentiate these t"o functions ,o difference in pitch patterns #-% range, -% movement' could be found bet"een response tokens "ith the t"o different conversational functions. pitch characteristics could vary bet"een e/tremes #eg rising and falling pitch movements' "ithin classes that "ere distinct from an interactional point of vie" 0isual cues such as gestures and ga1e did not predict these differences either Ho"ever, it "as observed that pitch characteristics of tokens of the same class appeared to depend on the pitch characteristics of the immediate prior talk of the interactional partner !f, for e/ample, the prior talk ended "ith rising intonation, the response token "as also produced "ith a rise if it "as encouraging the other speaker to continue talking !n order to perform the same action, the intonation of the 2uhu3 "as falling if the previous talk ended "ith a fall This "as not the case "hen the utterer of the 2uhu3 "as taking the floor or pro4ecting to do so !n order to 5uantify these effects, a techni5ue that ob4ectively measures the similarity of prosodic features such as pitch movement and individual speakers6 ranges is presented Using some e/amples "e e/plore ho" copying vs non(copying behaviour in the prosodic domain of short response tokens is used to manage talk in multi(party interaction #S1c1epek 7eed, $%%8' The analysis is based on the principal of cross(correlation "here the similarity of t"o signals 9 here -% contours 9 is established The method is e/tensible to cover other prosodic cues such as intensity or tempo, and even visual cues such as head nodding 7eferences. Gardner, 7 #$%%&' When listeners talk: response tokens and listener stance John +en4amins, )msterdam )*!(meeting corpus. http.::corpusamipro4ectorg: S1c1epek 7eed, + #$%%8' Prosodic Orientation in English Conversation ;algrave *ac*illan