| 摘要 |
PURPOSE:To allow continuous voice recognition with high reliability by correcting the recognition error of affiliated words from the sensible knowledge of verbs and the plural nouns which can be the case elements thereof among the results of the continuous voice recognition. CONSTITUTION:The information on independent words and affiliated words is sent from the information of an independent word dictionary 104 and an affiliated word dictionary 103 to an affiliated word correcting section 105 in accordance with the results of the continuous voice recognition by a voice recognition section 102. The fact that one of the two or other cases for the verb is the object thereof and the affiliated word connects thereto is known by using the meaning knowledge between the independent word and the affiliated word from a meaning knowledge section 106. Further, the fact that the sentence is unnatural is known and the recognized voice is corrected if the knowledge of a sensible knowledge memory section 107 is used. The recognition of the affiliated words which are relatively hardly recognizable than the independent words in the continuous voice recognition is executed with high accuracy in this way and the reliability of the recognition is enhanced. |
