Learning Pronunciation and Formulation Variants in Continuous Speech Applications

Abstract

Most voice driven applications are based on recognition grammars. In complex applications it is difficult to exactly predict how the users will formulate their requests even if a careful study of the user's behavior has been performed. Moreover, a speaker word pronunciation could not match well the phonetic transcription of the system, mainly in the case of foreign words. Loquendo has developed a tool that collects field data, detects the most significant weaknesses of the application due to pronunciation of formulation mismatches, and filters the collected field corpora, allowing the application designers to perform their analysis only on a reasonable amount of preprocessed and automatically labeled data. This paper presents the approaches that have been devised to detect vocabulary words pronunciation variants and linguistic formulations not covered by the recognition grammar. Results are given that show the improvements that have been obtained including automatically detected formulations in three grammars for two languages.

Oldenburg logatome speech corpus (OLLO) for speech recognition experiments with humans and machines

“…The emission probabilities of the HMM states were estimated by a Multi Layer Perceptron. Decoding of these states is carried out at phonetic level [8]. A phonetic transcription of logatomes was used to investigate the influence of variabilities.…”