METHOD AND APPARATUS FOR PROCESSING SPEECH RECOGNITION ERROR IN SPOKEN DIALOG SYSTEM, PARTICULARLY IN CONNECTION WITH IMPROVING SPEECH RECOGNITION RATE AND DIALOG SUCCESS RATE

摘要

PURPOSE: A method and an apparatus for processing a speech recognition error in a spoken dialog system are provided to improve a speech recognition rate and a dialog success rate by asking an ambiguous part to a speaker. CONSTITUTION: A speech recognition unit(120) extracts a characteristic of a speech signal inputted from a user, and finds out a phoneme closest to the characteristic to recognize words. A recognition error judging unit(130) obtains sentence reliability based on reliability of the recognized words, inspects a meaning structure of the sentence, and judges whether the recognized sentence has an error according to the sentence reliability and a result of the meaning structure inspection. A meta-dialog generating unit(140) generates a question to be asked to the user by considering contents where the error exists, and a kind of the error.