A Graphical Parametric Language-Independent Tool for the Annotation of Speech Corpora

Abstract

Robust speech recognizers and synthesizers require well-annotated corpora in order to be trained and tested, thus making speech annotation tools crucial in speech technology. It is very important that these tools are parametric so that they can handle various directory and file structures and deal with different waveform and transcription formats. They should also be language-independent, provide a user-friendly interface or even interact with other kinds of speech processing software. In this paper we describe an efficient tool able to cope with the above requirements. It was first developed for the annotation of the SpeechDat-II recordings, and then it was extended to incorporate the additional features of the SpeechDat-Car project. Nevertheless, it has been parameterized so that it is not restricted to the SpeechDat format and Greek, and it can handle any other formalism and language.

Resources

Details