Detection and recognition of correction utterances on misrecognition of spoken dialog system
✍ Scribed by Norihide Kitaoka; Naoko Kakutani; Seiichi Nakagawa
- Publisher
- John Wiley and Sons
- Year
- 2005
- Tongue
- English
- Weight
- 458 KB
- Volume
- 36
- Category
- Article
- ISSN
- 0882-1666
No coin nor oath required. For personal study only.
✦ Synopsis
Abstract
When we communicate with computers through a speech interface, misrecognition is inevitable. Detection of repetition by users is helpful for the system to detect the error and recover. In this paper, we assume that the utterance includes repetitions for correction and propose a method to detect correction utterances in spontaneously spoken dialog. We combined a word spotting method based on dynamic time warping (DTW) and an N‐best hypothesis overlap measure for the purpose. We obtained a recall rate of 92.7% and a precision of 89.1%. We next used the detection of correction utterances to improve recognition accuracy. © 2005 Wiley Periodicals, Inc. Syst Comp Jpn, 36(11): 24–33, 2005; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/scj.20341