Sixth International Conference on Spoken Language Processing
Techniques are investigated that use acoustic information from existing source language databases to implement automatic speech recognition (ASR) systems for new target languages for which little data are available. Strategies for cross-language use of acoustic information are proposed and are implemented via maximum a posteriori probability (MAP) and transformation-based techniques, as well as via discriminative learning techniques. The discriminative learning technique used is based on a cost-based extension of the minimum classification error (MCE) approach. Experiments are performed using relatively large amounts of English speech data from either a separate database or from the same database as smaller amounts of Afrikaans speech data to improve the performance of an Afrikaans speech recogniser. Results indicate that a significant reduction in word error rate is achievable (between 14% and 48% for experiments), depending on the method used and the amount of target language data available.
Bibliographic reference. Nieuwoudt, C. / Botha, Elizabeth C. (2000): "Cross-language use of acoustic information for automatic speech recognition", In ICSLP-2000, vol.3, 722-725.