IMPROVEMENT OF NEW WORD PRONUNCIATION STUDY BY USING PRONUNCIATION GRAPH

摘要

PPROBLEM TO BE SOLVED: To provide a method of converting word text and word pronunciation by a user into a phonographic description in order to add to a speech recognition lexicon, and a computer readable medium. PSOLUTION: At first, a plurality of possible phonographic descriptions are formed at least two. A phonographic description is formed by decoding the speech signal indicating the word pronunciation by a user. At least one of other phonographic descriptions is formed from a word text. A plurality of possible sequences including the phonographic description of a speech base and a text base are aligned, and scored in a single graph based on the correspondence between those and the pronunciation of the user. Next, the phonographic description having the highest score is selected for input into the speech recognition lexicon. PCOPYRIGHT: (C)2005,JPO&NCIPI