Syllabic consonants at different speaking rates: A problem for automatic speech recognition

Abstract A familiar aspect of English pronunciation is the occurrence of syllabic consonants. It is common to treat consonantal syllabicity as a consequence of vowel elision, implying that, for example, the pronunciation of “button” as /bλtn/ is the realisation of underlying /bλtən/ (or even /bλt⊃n/). Since elision is a phenomenon that is subject to the influence of speaking style, it would seem to follow that in rapid or casual speech we should expect to find more cases of syllabic consonants and fewer cases of unstressed vowels followed by continuant consonants. This paper sets out to show that the phenomenon is not this simple: we look at problems that confront our attempts at the automatic recognition of syllables and other sub-word units, and consider phonotactic and phonetic factors that may help to resolve them.