Abstract

It is widely acknowledged that pronunciations in spontaneous speech differ significantly from citation form. For this reason, pronunciation modeling has received considerable attention in recent automatic speech recognition literature. Most of the attention however has focussed on describing an alternate pronunciation as a different sequence of phonetic units using the same inventory of phones which describe canonical pronunciations. This approach assumes that the deviation from the canonical pronunciation can be represented in terms of complete changes from the base-form phonemes to the surface-form phones. In this paper, we investigate the properties of pronunciation change in conversational speech and demonstrate that most of the time the change is only partial; a phone is not completely deleted or substituted by another phone but it is modified only partially and the effects of this modification can be found in its environment. The evidence for this behavior comes from the analysis...

Citations