Method and apparatus for large vocabulary continuous speech recognition based on dynamic addition of vocabulary to be recognized

摘要

The present invention improves the recognition rate of the speech recognition system by using the multimedia data such as broadcasting, conference recording, call center conversation recording, and the like when users provide and add words or strings that are expected to be frequently generated errors. A method and apparatus for continuous speech recognition (large vocabulary continuous speech recognition) are proposed. According to an aspect of the present invention, in a continuous speech recognition system, applying a'BLANK' vocabulary to a language model in advance to construct a language model (vocabulary extension language model) capable of extending the vocabulary; Providing a user with a user-defined word/word phrase as an input to the speech recognition system, and inserting the user-defined word/word phrase into the'BLANK' vocabulary previously prepared in the vocabulary-extended language model search space; Provided is a method for continuously recognizing a large vocabulary based on a dynamic recognition target vocabulary, comprising receiving a user's spoken voice and performing a decoding process for the speech.