摘要 |
PURPOSE:To precisely detect the similarity and presence section of a language unit for a continuous voice like a sentence voice. CONSTITUTION:A pronunciation inscription processing part 8 outputs a phoneme inscription, generated by adding an adjacent possible phoneme to a phoneme inscription in a pronunciation dictionary 5, by referring an adjacent sound knowledge storage part 9 and a language unit similarity calculation part 6 after finding the similarity to the output of the pronunciation inscription processing part 8 traces back a calculation history to find the word similarity and the voice section of a target word. |