Interspeech'2005 - Eurospeech

Lisbon, Portugal
September 4-8, 2005

A Transformation-Based Learning Approach to Language Identification for Mixed-Lingual Text-to-Speech Synthesis

J. C. Marcadet (1), V. Fischer (2), C. Waast-Richard (1)

(1) IBM France, France; (2) IBM Deutschland Entwicklung GmbH, Germany

Recent progress in corpus-based concatenative text-to-speech synthesis has generated some interest in systems that are capable of synthesizing text from more than one language. In this paper we describe the language identification component of such a mixed-lingual text-to-speech system. Relying only on the input text, we employ two different methods, namely a transformation based learning approach and a stochastic n-gram approach, and we describe the combination of both methods. While the transformation-based learning approach already produces average error rates of less than 2 percent and outperforms the n-gram classification scheme, the combination of both methods results in a further error reduction of up to 50 percent.

Full Paper

Bibliographic reference.  Marcadet, J. C. / Fischer, V. / Waast-Richard, C. (2005): "A transformation-based learning approach to language identification for mixed-lingual text-to-speech synthesis", In INTERSPEECH-2005, 2249-2252.