Method for adapting to a speaker the phonetic transcriptions of a pronunciation lexicon

摘要

A method for adapting phonetic transcriptions of a pronunciation dictionary/lexicon specifically by taking into account the characteristics of a speaker or speakers, requires that the generated phoneme sequence is stored, in relation to a canonical form stored in the pronunciation lexicon as a K-P Tupel (sic), with reference to a speaker for each pronounced word, and the stored assignment is supplied to an input node of a neuronal network, as defined by the canonical form, and in the neuronal network the speaker adaptation is trained, in which the phoneme sequence is stipulated for the neuronal network as a target value for output nodes.