| 摘要 |
PURPOSE:To accurately set correct accents by utilizing the close relation among the phoneme array, phoneme feature and accent type of a borrowed word. CONSTITUTION:A control part 3 reads a character string out of an input character storage part 2, converts it into a Roman character string, and calculates the mora number of the word. Then the Roman character string is sent to a symbol pattern collating part 4, which collates the string with type patterns stored in a borrowed word type pattern and accent dictionary part 5. The borrowed word type pattern and accent dictionary indicates accent types by the mora positions of accent kernels counted from the end of the word. The symbol collating part 4 collates the input string with borrowed word type patterns in order and transfers accent position information to an output buffer 7 at the time of success. If, however, the collation is a failure, the control part 3 sends the Roman character string and mora number of the word to the symbol pattern collating part 4 again to collate them with symbol patterns in a phoneme pattern and accent dictionary part 6. Consequently, correct accents can accurately be set. |
