Timing in Speech: A Multi-Level Process

The modelling of timing in speech is of particular interest to language technology because it represents an interface between cognitive and mechanical aspects in the processes of speech production. Higher-level aspects of speech timing control are related to speaker-specific and utterance-contextspecific factors such that the same sequence of sounds produced by two different speakers (or by one speaker on two different occasions) will most likely have different timing characteristics. Lower level aspects that govern timing ensure on the other hand that there will also be some similarities in the productions since the vocal tract and articulatory mechanisms used to produce the individual sounds are fundamentally similar for all speakers of each language. The interaction of these two levels of influence results in patterns of timing that can be interpreted in a useful way to add an extra layer of meaning to an utterance.