The SPPAS Participation to the Forced-Alignment Task of Evalita 2011

摘要

SPPAS is a new tool to produce automatic annotations which include utterance, word, syllabic and phonemic segmentations from a recorded speech sound and its transcription. SPPAS is "user-friendly" and open source software issued under the GNU Public License. This paper describes SPPAS algorithms and Italian resources for phone-tization and alignment and evaluations related to the "Forced Alignment on Spontaneous Speech" task of the Evalita 2011 campaign. SPPAS is based on a dictionary look-ups approach for the phonetization and the use of the grammar-based Julius engine for alignment. A grammar contains sets of predefined combinations of words and contains one or more representations of the distinct phones that make up each word.