| 摘要 |
PURPOSE:To provide the sentence reading-out device which can register a correct accent, while using a voice dictionary for reading out a sentence without registering in advance accents of all words in the voice dictionary. CONSTITUTION:An input sentence in a step (hereinafter denoted as S) 1 is divided into words in S2, and an accent which is not registered yet in a dictionary is inferred in S3. In S4, pronunciation and an accent required for reading-out are generated, and in S5, a composite voice is outputted. In S6, user designates an erroneous accent phrase, and in S9, a correct accent is inputted. Based on the input accent in S9, an accent of an accent unregistered word is restored in S10, and in S12, the restored accent is registered in a dictionary. In the case the restoration fails in S10, the accent is registered together with the phrase in the dictionary as special information in S13. When the phrase is not designated by the user in S6, an inferred accent is registered in the dictionary in S8. |
