Summary: Lexical Access for Speech Understanding using
Minimum Message Length Encoding
Ian Thomas, Ingrid Zukerman, Jonathan Oliver, David Albrecht Bhavani Raskutti
Department of Computer Science Artificial Intelligence Section
Monash University Telstra Research Laboratories
Clayton, Victoria 3168, AUSTRALIA Clayton, Victoria 3168, AUSTRALIA
fiant,ingrid,jono,dwag@cs.monash.edu.au b.raskutti@trl.oz.au
Abstract
The Lexical Access Problem consists of deter­
mining the intended sequence of words corre­
sponding to an input sequence of phonemes (ba­
sic speech sounds) that come from a low­level
phoneme recognizer. In this paper we present
an information­theoretic approach based on the
Minimum Message Length Criterion for solv­
ing the Lexical Access Problem. We model
sentences using phoneme realizations seen in
training, and word and part­of­speech informa­
tion obtained from text corpora. We show res­
ults on multiple­speaker, continuous, read speech

Source: Albrecht, David - Caulfield School of Information Technology, Monash University

Collections: Computer Technologies and Information Sciences