9th Annual Conference of the International Speech Communication Association

Brisbane, Australia
September 22-26, 2008

Recent Improvements of the RWTH GALE Mandarin LVCSR System

Ch. Plahl (1), Björn Hoffmeister (1), M.-Y. Hwang (2), D. Lu (3), Georg Heigold (1), Jonas Loof (1), Ralf Schlüter (1), Hermann Ney (1)

(1) RWTH Aachen University, Germany
(2) Microsoft Research, USA
(3) Southwest Forestry University, China

This paper describes the current improvements of the RWTH Mandarin LVCSR system. We introduce a new reduced toneme set developed at RWTH. We are using different toneme sets and pronunciation lexica. For the purpose of discriminative training we will show a fast way to transform word lattices between systems using different toneme sets and pronunciation lexica. In addition to various acoustic front-ends, the current systems use different kinds of neural network toneme posterior features. While different kinds of systems are developed, a two stage decoding framework for combining these systems is applied. We show detailed recognition results of the development cycle of the systems. Finally, two methods to integrate tonal features are compared.

Full Paper

Bibliographic reference.  Plahl, Ch. / Hoffmeister, Björn / Hwang, M.-Y. / Lu, D. / Heigold, Georg / Loof, Jonas / Schlüter, Ralf / Ney, Hermann (2008): "Recent improvements of the RWTH GALE Mandarin LVCSR system", In INTERSPEECH-2008, 2426-2429.