METHOD FOR AUTOMATIC SPEECH RECOGNITION OF ARBITRARY SPOKEN WORDS

A method and system for allowing an automatic speech recognition (ASR) system to recognize arbitrary words, by accessing information in a supplemental data base. A supplemental data base is accessed to retrieve supplementary textual information, such as a proper name. A text-to-speech means is used to generate a phoneme transcription of the text retrieved from the supplemental data base, so that the transcription may be used as a speaker independent template by the ASR system for recognizing a spoken word.