TRAED: Speech Audio Editing using Imperfect Transcripts

摘要

Although digital recording of speech is widespread, and an increasing range of applications allow recording and inclusion of speech data in documents, editing and retrieval of speech audio remains generally a challenging task. We have previously developed a speech audio editing and browsing application which utilizes imperfect transcripts of speech as a mechanism for text-based editing and retrieval of speech audio documents. This paper presents a second prototype, called TRAED, which enhances the functionality provided by our earlier prototype, and further facilitates the task of speech audio editing and access.