Predicting children’s hyperarticulate speech during human-computer error resolution
Sharon L. Oviatt, Rachel Coulston, Courtney Darves · The Journal of the Acoustical Society of America · 2003
When speaking to interactive systems, people sometimes hyperarticulate—or adopt a clarified form of speech that has been associated with increased recognition errors. The goal of the present study was to provide a comprehensive assessment of the type and magnitude of linguistic adaptations in children’s speech during human-computer error resolution, and to compare these adaptations with those typical of adult hyperarticulation. A study was conducted in which twenty-four 7- to- 10-year-old children interacted with a simulated conversational system, which permitted a comparison of their verbatim repetitions immediately before and after system recognition errors. Matched original-repeat utterance pairs then were analyzed for acoustic, prosodic, and phonological adaptations. Like adult speech, the primary hyperarticulate changes in children’s speech included durational phenomena such as lengthening of pauses and the speech segment, and a more deliberate, hyper-clear articulatory style. However, children’s speech also displayed large increases in amplitude that are not typical of adult hyperarticulation, as well as substantially larger magnitude adaptations than those observed in adult speech. These results corroborate and generalize the Computer-elicited Hyperarticulate Adaptation Model, and have implications for improved error handling in next-generation spoken language and multimodal systems. [Work supported by NSF Grant No. IIS-0117868.]