- Results from research performed on small, controlled extracts of speech
cannot be generalised to describe all speech. This is important for
both basic research and speech technology.
- A variety of different speech corpora are already available.
- A well-annotated corpus enables the researcher to explore the phenomena
under study in greater detail, to try out different possibilities,
and to see the extent of variability more clearly.
- Language technology usually requires large quantities of speech material
(development and testing of automatic speech recognizers, exemplars
for speech synthesis)