Processing Japanese Self-correction in Speech Dialog Systems.

Abstract

Speech dialog systems need to deal with various kinds of ill-formed speech inputs that appear in natural human-human dialog. Self-correction (or speech-repair) is a particularly problematic phenomenon. Although many w ays of dealing with self-correction have been proposed, these have limitations in both detecting and correcting for this phenomenon. In this paper, we propose a method to overcome these problems in Japanese speech dialog. We evaluate the proposed method using our speech dialog corpus and discuss its limitations and the work that remains to be done.