Sixth International Conference on Spoken Language Processing
In this era of globalization, learning multiple languages is becoming necessary. Learning to speak a new language involves knowing how to correctly pronounce words. In many cases, corrections to pronunciation mistakes come from language teachers who can typically give students only limited time and attention. With the rapid development of automatic speech recognition (ASR) technologies, computer can now accurately transcribe spoken words.
In this paper, we reported our work of using ASR to help students differentiate between comfusible word pairs, commonly known as minimal pairs. Most of the reported work focuses on the phonetic level. However, in Mandarin, we have to deal with the added dimension of tonal confusion where two words differ only by their lexical tones. We found that tone is consistently harder to distinguish as compared to sylable initial or final. Furthermore, the system when tested on native system can achieve an accuracy of over 90%
Bibliographic reference. Siu, Man-hung / Wong, Ka-Ming / Ching, Man-Yan / Lau, Mei-Sum (2000): "Computer-aided Mandarin pronunciation learning system", In ICSLP-2000, vol.4, 77-80.