UPDATING METHOD FOR LANGUAGE MODEL OF SPEECH APPLICATION

摘要

PURPOSE: An updating method is provided to automatically update the language model of a speech recognition engine in a speech application for the session of correcting the wrong recognition of speech without calling the dialog box of a user. CONSTITUTION: A sound of an initially dictated text and the sound of an alternative text are automatically compared. If the sounds are similar for the alternative text to represent the correction of wrong recognition rather than an edit, a language model is automatically updated by correction. Therefore, a new method for updating the language model of a speech recognition engine in a speech application for sessions, which correct the wrong recognition of speech, does not requires to call the dialogue box of a user interaction.