APPARATUS AND METHOD FOR ADDING NEW PROPER NOUNS TO LANGUAGE MODEL IN A CONTINUOUS SPEECH RECOGNITION SYSTEM

摘要

PURPOSE: A device for adding a proper noun of a language model in a continuous speech recognition system and a method thereof are provided to obtain a high voice recognition performance of the proper noun by adding a new proper noun in an n-gram to the language model. CONSTITUTION: A collecting unit(102) collects classification words and a new proper noun from a proper noun addition unit of a language model. A candidate sentence selecting unit(106) selects a candidate sentence. A candidate framework extracting unit(110) extracts a candidate framework from the candidate sentence. A candidate framework generating unit(112) of an n-gram extracts a candidate framework in an n-gram type from the candidate framework. An n-gram extending unit(116) substitutes the collected proper noun for the candidate framework in the n-gram type. A reflecting unit(118) adds the generated n-gram to the language model. [Reference numerals] (102) Collecting unit; (104) Training text focus; (106) Candidate sentence selecting unit; (108) Text focus; (110) Candidate framework extracting unit; (112) n-gram candidate framework generating unit; (114) Ranking determining unit; (116) n-gram extending unit; (118) Reflecting unit; (120) Language model