Modeling Pronunciation Variation for Automatic Speech Recognition Rolduc, The Netherlands May 4-6, 1998

Joint Acoustic Unit Design and Lexicon Generation

M. Bacchiani, Mari Ostendorf

Electrical and Computer Engineering Department, Boston University, Boston, MA, USA

Although most parameters in a speech recognition system are estimated from data by use of an objective function, the unit inventory and lexicon are generally hand crafted and therefore unlikely to be optimal. This paper proposes a joint solution to the related problems of learning a unit inventory and corresponding lexicon from data. The proposed algorithm performs comparably to a state-of-the-art phone-based system on a speaker independent read speech task with moderate vocabulary size.

Full Paper

Bibliographic reference.  Bacchiani, M. / Ostendorf, Mari (1998): "Joint acoustic unit design and lexicon generation", In MPV-1998, 7-12.