| 摘要 |
PURPOSE:To successively and automatically bring a recognition candidate in a voice section to menu display by executing a language processing by using a recognition result from a recognizing part and generating a grammatically correct candidate train by a word unit or a sentence clause unit. CONSTITUTION:Based on a feature parameter generated by a voice analyzing part 1, whether a silent section of an input voice is a word boundary or a sentence clause boundary is discriminated in accordance with a prescribed rule by a boundary detecting part 4, and the word boundary or the sentence clause boundary is detected. Subsequently, with respect to a voice section which is inputted until the time point when a word boundary or a sentence clause boundary is detected by a boundary detecting part 4, and also, in which a recognition result is not defined yet, a language processing is executed successively and automatically by a language processing part by using a recognition result from a recognizing part 3, and grammatically correct candidate train is generated by a word unit or a sentence clause unit. Also, this generated candidate train is inputted to a display part 7, and this candidate train is displayed by menu method. In such a way, a candidate of a voice section inputted until its time point is brought to menu display. |
