Paper: Predicting Barge-in Utterance Errors by using Implicitly-Supervised ASR Accuracy and Barge-in Rate per User

ACL ID P09-2023
Title Predicting Barge-in Utterance Errors by using Implicitly-Supervised ASR Accuracy and Barge-in Rate per User
Venue Annual Meeting of the Association of Computational Linguistics
Session Short Paper
Year 2009
Authors

Modeling of individual users is a promis- ing way of improving the performance of spoken dialogue systems deployed for the general public and utilized repeatedly. We define “implicitly-supervised” ASR accu- racy per user on the basis of responses following the system’s explicit confirma- tions. We combine the estimated ASR ac- curacy with the user’s barge-in rate, which represents how well the user is accus- tomed to using the system, to predict in- terpretation errors in barge-in utterances. Experimental results showed that the es- timated ASR accuracy improved predic- tion performance. Since this ASR accu- racy and the barge-in rate are obtainable at runtime, they improve prediction perfor- mance without the need for manual label- ing.