| 摘要 |
PROBLEM TO BE SOLVED: To provide an application apparatus of phonetic transcription in Kana, and program thereof, which dispense with computation of unrealistic phonetic transcription candidates in Kana, thereby improving efficiency and convenience thereof. SOLUTION: In the application apparatus for phonetic transcription in Kana 50, a dictionary-registering pronunciation of a word (information of a part of speech is not necessary), and a language model, that discriminates and provides conjunctive probability between the words having the same notation and different pronunciation are arranged. An assumption generation/deletion part 30 generates an assumption for the phonetic transcription candidates in Kana for a Japanese text by the dictionary. An assumption computing part 20 computes the product of the conjunctive probability, between the words in the assumption on the basis of the language model, merges so as to leave only the assumption where the product of the conjunctive probability between the words becomes maximum, selects the assumption where the product of the conjunctive probability between the words at a sentence end becomes maximum and outputs a pronunciation string. |
