×

Method and apparatus handling speech recognition errors in spoken dialogue systems

  • US 7,493,257 B2
  • Filed: 08/05/2004
  • Issued: 02/17/2009
  • Est. Priority Date: 08/06/2003
  • Status: Active Grant
First Claim
Patent Images

1. An apparatus handling speech recognition errors in a spoken dialogue system, comprising:

  • a speech recognition unit extracting a speech feature of a speech signal inputted from a user and finding a phoneme nearest to the speech feature to recognize a word;

    a recognition error determination unit finding a sentence confidence based on a confidence of the recognized word, inspecting a semantic structure of a recognized sentence, and determining whether an error exists in the recognized sentence which is subject to speech recognition according to a predetermined criterion based on both the sentence confidence and a result of inspecting the semantic structure of the recognized sentence; and

    a meta-dialogue generation unit generating a question questioning the user about the recognized sentence based on a content of a portion where the error exists and a type of the errors,wherein the question is adapted so that, after both a confidence θ

    2 determined that each word is a correct answer and another confidence θ

    1 determined that each word is an incorrect answer are obtained, the word is substituted by an interrogative when the confidence of each recognized word is less than θ

    1, the word is omitted without repetition when the confidence of each recognized word is greater than θ

    2, and a phrase of the recognized word is repeatedly questioned when the confidence of each recognized word is greater than or equal to θ

    1, and is less than or equal to θ

    2.

View all claims
  • 1 Assignment
Timeline View
Assignment View
    ×
    ×