CONSIDERING THE PECULIARITY OF SPONTANEOUS SPEECH AT THE DEVELOPING AUTOMATIC SPEECH RECOGNITION SYSTEM
I. S. Kipyatkova
St. Petersburg Institute for Informatics and Automation of the Russian Academy of Sciences (SPIIRAS), Saint Petersburg, 199178, Russian Federation; senior researcher
Abstract. An approach to automatic processing of spontaneous Russian speech, consisting in recognition of nonphonemic elements and modeling of alternative variants of word pronunciation, is considered. A set of acoustic and lexical models, that are intended for separation of noises from keywords and that take into account possible elements of spontaneous speech, is represented. An algorithm of creation of alternative transcriptions by extended transcribing rules is presented. Experimental results are represente
Keywords: speech recognition, alternative transcriptions, non-phonemic elements.