APPARATUS AND METHOD FOR GENERATING MULTIPLE PRONUNCIATION DICTIONARY BASED ON STATISTICS

摘要

PURPOSE: A multi pronunciation dictionary generating device and a method thereof are provided to use source information and calculated frequency by calculating the frequency of candidate pronunciations of vocabularies appearing in a mass voice database and storing the source information. CONSTITUTION: A voice-pronunciation arranging unit(120) detects a pronunciation the closest to a word included in a voice signal file(111) from a multi pronunciation dictionary by using an arrangement function of a voice recognizer. A word-pronunciation pair extracting unit(140) applies the detection of the closest pronunciation to a word level document(112) and voice signal files stored in a database(110). A pronunciation statistics information extracting unit(150) calculates statistics information about pronunciations of each word in the multi pronunciation dictionary based on the extracted word and pairs of the pronunciations. [Reference numerals] (110) Database; (111) Voice signal file; (112) Word level document; (113) Speaker information; (120) Voice-pronunciation arranging unit; (130) Multiple pronunciation dictionary; (140) Word-pronunciation pair extracting unit; (150) Pronunciation statistics information extracting unit; (160) Hash key value generating unit; (170) Duplication using verification unit; (180) Statistics-based multiple pronunciation dictionary