| 摘要 |
PROBLEM TO BE SOLVED: To synthesize voices with easily understandable pronunciation approximate to Japanese with the same speaker as Japanese voices when the words of foreign language notation coexist in Japanese text by discriminating and assorting the foreign language portions in an input character system, converting the assorted portions to a pronunciation symbol system and converting this pronunciation symbol system into a KANA (Japanese syllabary) system. SOLUTION: The foreign language parts in the input characteristic strings of Japanese are discriminated by a language discrimination section 1. The portions of Japanese mainly composed of the text in which KANJI (Chinese character) and KANA are mixed, are determined of reading of KANJI by processing in a morpheme analysis section 4, are made into a reading KANA system and are converted to continuous speeches in a Japanese speech synthesis section 8. The portions decided to be foreign language are sent to a pronunciation symbol string conversion section 6. When the pronunciation symbol strings are obtained, the pronunciation symbol strings are converted to KANA in a KANA conversion section 7 by the corresponding relation with the Japanese speech pronunciation symbols and are converted to the continuous speeches in the Japanese speech synthesis section 8. Accordingly, the synthesized tones extremely easily hearable to Japanese can be obtained for the expression with the pronunciation like that for the words of foreign origin. |
