Active learning for spoken language understanding
First Claim
1. A method generating a classifier from training data St and a larger amount of unlabeled data in a pool Su, the method comprising:
- (1) training a classifier using current training data St, the training data St generated by sampling a plurality of utterances;
(2) classifying utterances in a pool Su using the trained classifier;
(3) computing a call type confidence score for each utterance;
(4) sorting candidate utterances with respect to the confidence score of the maximum scoring call type;
(5) selecting the lowest scored k utterances from Su using the confidence scores and labeling them to define a labeled set Si;
(6) redefining St=St∪
Si; and
(7) redefining Su=Su−
Si.
5 Assignments
0 Petitions
Accused Products
Abstract
Disclosed is a system and method of training a spoken language understanding module. Such a module may be utilized in a spoken dialog system. The method of training a spoken language understanding module comprises training acoustic and language models using a small set of transcribed data ST, recognizing utterances in a set Su that are candidates for transcription using the acoustic and language models, computing confidence scores of the utterances, selecting k utterances that have the smallest confidence scores from Su and transcribing them into a new set Si, redefining St as the union of St and Si, redefining Su as Su minus Si, and returning to the step of training acoustic and language models if word accuracy has not converged.
-
Citations
16 Claims
-
1. A method generating a classifier from training data St and a larger amount of unlabeled data in a pool Su, the method comprising:
-
(1) training a classifier using current training data St, the training data St generated by sampling a plurality of utterances; (2) classifying utterances in a pool Su using the trained classifier; (3) computing a call type confidence score for each utterance; (4) sorting candidate utterances with respect to the confidence score of the maximum scoring call type; (5) selecting the lowest scored k utterances from Su using the confidence scores and labeling them to define a labeled set Si; (6) redefining St=St∪
Si; and(7) redefining Su=Su−
Si. - View Dependent Claims (2, 3, 4, 5, 6)
-
-
7. A system having spoken language understanding module generated according to a method comprising:
-
(1) training a plurality of classifiers independently using a training data set St, the training data St generated by sampling a plurality of utterances; (2) classifying utterances in a set Su using the plurality of classifiers and computing a call type confidence score for all utterances; (3) sorting candidate utterances with respect to a score of the maximum scoring call type according to one of the classifiers if the classifiers disagree; (4) selecting and labeling the lowest scored k utterances from Su to define a labeled set Si and redefining St and Su as follows; (5) St=St∪
Si; and(6) Su=Su−
St, wherein the labeled utterances are used to generate the spoken language understanding module. - View Dependent Claims (8, 9)
-
-
10. A system having spoken language understanding module trained using a method comprising:
-
(1) training acoustic and language models using a small set of transcribed data St, the training data St generated by sampling a plurality of utterances; (2) recognizing utterances in a set Su that are candidates for transcription using the acoustic and language models; (3) computing confidence scores of the utterances; (4) selecting the lowest scored k utterances from Su using the confidence scores and transcribing them into a new set Si; (5) redefining St as the union of St and Si; (6) redefining Su as Su minus Si; and (7) returning to step (1) of word accuracy has not converged. - View Dependent Claims (11, 12)
-
-
13. A method of generating a spoken language understanding module, the method comprising, from small amount of training data St and a larger amount of unlabeled data Su:
-
(1) training a plurality a classifiers independently using a training data set Si, the training data St generated by sampling a plurality of utterances; (2) classifying utterances in a set Su using the plurality of classifiers and computing a call type confidence score for all utterances; (3) sorting candidate utterances with respect to a score of the maximum scoring call type according to one of the classifiers if the classifiers disagree; (4) selecting and labeling the lowest scored k utterances from Su to define a labeled set Si and redefining St and Su as follows; (5) St=St∪
Si; and(6) Su=Su−
Si, wherein the labeled utterances are used to generate the spoken language understanding module. - View Dependent Claims (14, 15, 16)
-
Specification