| 摘要 |
PURPOSE:To convert recognized sounds into a ''kanji'' (Chinese character)/''kana'' (Japanese syllabary) mixed sentense with a high conversion ratio by executing the automatic correcting processing for misrecognition of a recognized syllable. CONSTITUTION:All combinations of an input syllable matrix including proposed syllables are compared with an independent word index dictionary to extract the proposed reading ''kana''s of an independent word and to obtain the conjugation information for the parts of speech of the reading ''kana''. Then, all the combinations of the input syllable matrix excluding the independent words are compared with a conjugated ending directionary and an adjunct dictionary to extract conjugated endings and adjuncts. Then, the priority of clause reading ''kana''s to execute ''kana''/''kanji'' conversion while considering the grammatical connection and recognition ratio is determined, then the plainly determined clause reading ''kana''s are converted into a ''kanji''/''kana'' mixed sentense. Automatic correction for misrecognition of adjuncts in adjunct reading ''kana'' extracting processing uses characteristics that the misrecognition of a consonant part of a single syllable may exist, but the misrecognition of a vowel part does not exist.
|
