JAPANESE SPEECH RECOGNIZING METHOD

摘要

PURPOSE:To provide the Japanese speech recognizing method which recognizes an optional speech fast efficiently. CONSTITUTION:A statistical language model 21 regarding the appearance order of KANA(Japanese syllabary) and KANJI(Chinese character) is generated (step 100) as a statistical language model from a learning text data base 20, new text data different from the learning text data base 20 is automatically generated (step 101) by using the learning text data base 20 and statistical language model 21, and the recognition search range of the phoneme series of the inputted speech is restricted (step 102) by referring to the learning text data base and new text data.