A learning system for a more complete process of learning to read aloud English words should consist of three major steps: alignment, mapping learning, and grapheme generation. In a previous paper, we discussed a basi alignment algorithm and several improvements that effectively accomplish the alignment and mapping learning simulatneously. In this paper, we focus mainly on grapheme generation. We present several criteria for constructing graphemes, and demonstrate that graphemes generated improve the predictive accuracy in learning to read aloud.
展开▼