What makes speech data spontaneous?
Daniela Oppermann, Susanne Burger · 1999
The aim of the work to be reported here is the development of schemata which are able to predict the quality of spontaneity and help to create and collect databases for certain tasks on spontaneous speech. The term "spontaneous speech" is used in a wide range and allows the existence of many spontaneous speech corpora with different levels of spontaneity. Our aim was to find appropriate categories and description values for these corpora. Therefore we started by analyzing transcriptions of spontaneous monologues of one minute, which were recorded and annotated. We made a structure analysis of the introductory part of the monologues and let people qualify categories of spontaneity in a small experiment containing a subset of the monologues. Correlations between the judged categories of spontaneity and the amount of spontaneous speech phenomena in the monologues will be shown. 1. INTRODUCTION In recent years speech recognition has concentrated more and more on understanding spontaneous...