Algorithms for syllabic hypothesization in continuous speech

摘要

The paper describes a new method for representing and using syllabic knowledge in a Speech Understanding System. The aims of this representation are the generation of syllabic hypotheses, given in terms of phonetic features and phonemes, (KS used in a data-driven way) and the better specification of consonant clusters (KS used in a model-driven way). The generation of syllabic hypotheses is performed by a non-directional parser.