Speech recognition vocabulary integration for classifying words to identify vocabulary application group

Patent metadata

Field	Value
Publication number	US-9305545-B2
Application number	US-201313802390-A
Country	US
Kind code	B2
Filing date	Mar 13, 2013
Priority date	Mar 13, 2013
Publication date	Apr 5, 2016
Grant date	Apr 5, 2016

Abstract

A method for vocabulary integration of speech recognition comprises converting multiple speech signals into multiple words using a processor, applying confidence scores to the multiple words, classifying the multiple words into a plurality of classifications based on classification criteria and the confidence score for each word, determining if one or more of the multiple words are unrecognized based on the plurality of classifications, classifying each unrecognized word and detecting a match for the unrecognized word based on additional classification criteria, and upon detecting a match for an unrecognized word, converting at least a portion of the multiple speech signals corresponding to the unrecognized word into words.

First claim

What is claimed is:

1. A method for vocabulary integration in speech recognition, comprising:
converting a plurality of speech signals into a plurality of words using a processor;
applying a plurality of confidence scores to the plurality of words;
receiving classification criteria for a primary application;
classifying the plurality of words into a plurality of classifications to identify a vocabulary application group that the plurality of words belongs to by performing a first classification process based on using the classification criteria for the primary application and the confidence score for each one of the plurality of words;
determining one or more unrecognized words of the plurality of words based on the plurality of classifications;
receiving additional classification criteria for multiple secondary vocabulary applications for the one or more unrecognized words, wherein the multiple secondary vocabulary applications comprise application specific vocabularies;
classifying each of the one or more unrecognized words to identify a vocabulary application group that the one or more unrecognized words belong to by performing a second classification process based on using the received additional classification criteria and confidence scores determined for the one or more unrecognized words;
detecting a match for each of the one or more unrecognized words based on the classifying each of the one or more unrecognized words with the received additional classification criteria; and
upon detecting a match for an unrecognized word, converting at least a portion of the plurality of speech signals corresponding to the unrecognized word into a plurality of recognized words.

2. The method of claim 1 , further comprising:
receiving the plurality of speech signals using an electronic device, wherein a primary vocabulary application is used for converting the plurality of speech signals into the plurality of words; and
receiving the additional classification criteria for the multiple secondary vocabulary applications for the one or more unrecognized words, wherein the multiple application specific vocabularies include application specific terms that are integrated with terms from the primary vocabulary application into an automatic speech recognition (ASR) engine.

3. The method of claim 2 , wherein the primary vocabulary application is a generic vocabulary application, and the multiple secondary vocabulary applications are third-party provided vocabulary applications, and each of the third-party provided vocabulary applications comprise vocabulary that is specific to a third-party provider.

4. The method of claim 3 , wherein classifying the plurality of words and classifying the one or more unrecognized words using the first classification process and the second classification process are based on using a classifier comprising at least one of binary classification, use of feature vectors, linear classification and non-linear classification.

5. The method of claim 4 , further comprising:
converting all of the plurality of speech signals to words using the specific one of the multiple secondary vocabulary applications.

6. The method of claim 5 , further comprising:
using a natural language understanding process for recognizing the converted words.

7. The method of claim 6 , wherein the third-party vocabulary applications comprise one or more vocabulary lists.

8. The method of claim 7 , wherein the one or more vocabulary lists are provided one at a time for subsequent speech recognition processes.

9. The method of claim 2 , wherein converting the plurality of speech signals to words based on using the primary vocabulary application comprises a first speech recognition process; and converting at least the portion of the plurality of speech signals to words using the specific one of the multiple secondary vocabulary applications comprises a second speech recognition process.

10. The method of claim 9 , wherein the first speech recognition process and the second speech recognition process are both performed together, and a result of the first speech recognition process and a result of the second speech recognition process combine to provide a recognized speech result.

11. The method of claim 1 , wherein the application specific vocabularies comprise made-up words.

12. The method of claim 1 , further comprising:
requesting the multiple secondary vocabularies that are limited to one or more application specific vocabularies that are classified as having a possible match for the one or more unrecognized words.

13. A system for vocabulary integration of speech recognition, comprising:
an electronic device including a microphone configured to receive a plurality of speech signals; and
an automatic speech recognition (ASR) engine configured to convert the plurality of speech signals into a plurality of words, the ASR engine comprising:
a vocabulary application interface configured to integrate multiple vocabulary applications; and
a classifier configured to provide classification for a primary vocabulary application to identify a vocabulary application group that the plurality of words belongs to by performing a first classification process based on using classification criteria and confidence scores determined for the plurality of words, and to provide classification for one or more unrecognized words to identify a vocabulary application group that the one or more unrecognized words belongs to by performing a second classification process based on additional classification criteria for multiple secondary vocabulary applications and confidence scores determined for the one or more unrecognized words, wherein the multiple secondary vocabulary applications comprise application specific vocabularies,
and a plurality of unrecognized words resulting from a first speech recognition pass using the ASR engine are converted into a plurality of recognized words using the ASR engine for a second speech recognition pass for providing a recognized speech result.

14. The system of claim 13 , wherein the ASR engine is configured to convert the plurality of speech signals into the plurality of words based on using the primary vocabulary application and the multiple secondary vocabulary applications.

15. The system of claim 14 , wherein the primary vocabulary application is a generic vocabulary application, and the multiple secondary vocabulary applications are third-party provided vocabulary applications, the first speech recognition pass is based on the primary vocabulary application and the second speech recognition pass is based on the multiple secondary vocabulary applications, and each of the third-party provided vocabulary applications comprise vocabulary that is specific to a third-party provider.

16. The system of claim 15 , wherein the classifier is configured to classify recognized and unrecognized words using the first classification process and the second classification process based on employing a processor for performing at least one of binary classification, use of feature vectors, linear classification and non-linear classification.

17. The system of claim 15 , wherein the third-party vocabulary applications comprise one or more vocabulary lists, and a third-party application comprising multiple vocabulary lists provides each vocabulary list one at a time for subsequent speech recognition processes.

18. The system of claim 15 , wherein the electronic device is a mobile phone.

19. The system of claim 13 , wherein the application specific voc

Assignees

• Samsung Electronics Co Ltd

Inventors

• Cheung Chun Shing

Classifications

• G10L15/20

  Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech (G10L21/02 takes precedence) · CPC title

• G10L15/18 Primary

  using natural language modelling · CPC title

• G10L15/08

  Speech classification or search · CPC title

• G10L15/26 Primary

  Speech to text systems (G10L15/08 takes precedence) · CPC title