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 repair) is a particularly problematic phenomenon. Although many methods of dealing with self-correction have been proposed, they have limitations in both detecting and correcting this phenomenon. In this paper, we propose a new method overcoming ill-formedness of speech inputs. We evaluate the proposed method using a speech dialog corpus and discuss its effectiveness and limitation.
