| 摘要 |
PROBLEM TO BE SOLVED: To solve the problem that in a conventional method, it takes much calculation time to attach read candidates and determine a segment position by performing morphological analysis on a text data, and that segmentation and read attachment of a description text data cause calculation time increase and recognition accuracy deterioration, as a plurality of read candidates are given for one description, and that segmentation and read attachment of a read text data cause segment position accuracy deterioration. SOLUTION: In order to create a vocabulary dictionary for speech recognition for recognizing any of text data of twinned description and reading of a description and reading text storage section by using the vocabulary dictionary for speech recognition, correspondence of the twinned description and reading of the text data from the description and reading text storage section, is determined by the description and reading correspondence means by referring to description and its reading candidates of the description and reading correspondence storage means. The text data of the corresponded description and reading are segmented by a segmentation means to a word or a phrase unit, and the vocabulary dictionary for speech recognition is created for the twinned description and reading after segmentation by a recognition vocabulary dictionary creation means, and stored in a recognition vocabulary dictionary storage section. COPYRIGHT: (C)2010,JPO&INPIT
|
