APPARATUS AND A METHOD FOR RECOGNIZING SPEECH BY ANALYZING USE PATTERN OF EACH USER TO IMPROVE A RECOGNITION RATIO OF NAMES THAT CAUSE LOWERING OF THE RECOGNITION RATIO

摘要

PURPOSE: An apparatus and a method for recognizing speech are provided to improve a recognition ratio of names, such as a homonym and a similar name, that cause lowering of the recognition ratio by analyzing characteristics of a use pattern by each user of a speech recognition service. CONSTITUTION: A preprocess part(11) analyzes a caller telephone number and then loads PNM(Personal Name Management) DB(Database). A speech recognition part(12) recognizes caller's speech and then determines a recognition result corresponding to a recognized word. A recognized-word determination part(13) rearranges the recognition result by utilizing the PNM DB and a recognized-word determination rule and thereby determines a final recognized word. A PNM DB management part(14) registers and deletes a name in and from an exception word list included in the PNM DB according to recognition success or failure. The PNM DB management part stores data of previous recognition success and failure by each caller telephone number in the PNM DB.