Method and apparatus handling speech recognition errors in spoken dialogue systems
First Claim
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.
1 Assignment
0 Petitions
Accused Products
Abstract
To handle portions of a recognized sentence having an error, a user is questioned about contents associated with portions. According to a user'"'"'s answer, a result is obtained. Speech recognition unit extracts a speech feature of a speech signal inputted from user and finds a phoneme nearest to the speech feature to recognize a word. Recognition error determination unit finds a sentence confidence based on a confidence of the recognized word, performs examination of a semantic structure of a recognized sentence, and determines whether or not an error exists in the recognized sentence which is subjected to speech recognition according to predetermined criterion based on both sentence confidence and result of examining semantic structure. Meta-dialogue generation unit generates a question asking user for additional information based on content of a portion where the error exists and a type of the error.
40 Citations
33 Claims
-
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 Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14)
-
-
15. A method of handling speech recognition errors in a spoken dialogue system, comprising:
-
extracting a speech feature of a speech signal inputted by a user and finding a phoneme nearest to the speech feature to recognize a word; determining whether an error exists in a sentence which is subject to speech recognition according to a predetermined criterion based on both a sentence confidence obtained from a confidence of the recognized word and a result of examining a semantic structure of the sentence; and generating a question questioning the user about the recognized sentence based on both 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 Dependent Claims (16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28)
-
-
29. An apparatus handling speech recognition errors, comprising:
-
a speech signal input unit receiving a speech signal input from a user; a speech recognition unit extracting a speech feature from the speech signal input from the user and recognizing words, generating a plurality of candidate words and forming a recognized sentence; a recognition error determination unit determining whether there is an error in the recognized sentence according to a predetermined criterion based on sentence confidence and semantic structure of the recognized sentence; and a meta-dialogue generation unit generating a question questioning the user about the recognized sentence based on both a content of a portion where the error exists and a type of the error, 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 Dependent Claims (30, 31)
-
-
32. A computer readable medium encoded with processing instructions performing a method of handling speech recognition errors in a spoken dialogue system, the method comprising:
-
extracting a speech feature of a speech signal inputted by a user and finding a phoneme nearest to the speech feature to recognize a word and generating a recognized sentence; determining whether an error exists in the recognized sentence according to a predetermined criterion based on both a sentence confidence obtained from a confidence of the recognized word and a result of examining a semantic structure of the recognized sentence; and generating a question questioning the user about the recognized sentence based on both a content of a portion where the error exists and a type of the error, 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 Dependent Claims (33)
-
Specification