Approach to reducing the response time of a speech interface

• US 9,922,647 B2
• Filed: 01/29/2016
• Issued: 03/20/2018
• Est. Priority Date: 01/29/2016
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Kurata, Gakuto, Nagano, Tohru
• Primary Examiner(s): Chawan, Vijay B
• Application Number:
  US15/010,725
• Publication Number:
  US 20170221486A1
• Time in Patent Office: 781 Days
• Field of Search: 704240, 704257, 704 10, 704 9, 704243, 704251, 704252, 704253, 704254, 704260, 704275, 707694, 707706, 707713, 707759, 707769, 709206, 709207
• CPC Class Codes: G10L 13/00 Speech synthesis; Text to speech systems
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/28 Constructional details of speech recognition systems

Abstract

A method for reducing response time in a speech interface including constructing a partially completed word sequence from a partially received utterance from a speaker received by an audio sensor, modeling a remainder portion using a processor based on a rich predictive model to predict the remainder portion, and responding to the partially completed word sequence and the predicted remainder portion using a natural language vocalization generator with a vocalization, wherein the vocalization is prepared before a complete utterance is received from the speaker and conveyed to the speaker by an audio transducer.

Claims

1. A method for reducing response time in a speech interface comprising:

constructing a partially completed word sequence from a partially received utterance from a speaker received by an audio sensor;

modeling a remainder portion for the partially received utterance using a processor based on a rich predictive model to predict the remainder portion; and

responding to the partially completed word sequence and the predicted remainder portion for the partially received utterance using a natural language vocalization generator with a vocalization, wherein the vocalization is prepared before a complete utterance is received from the speaker and conveyed to the speaker by an audio transducer.

2. The method according to claim 1, wherein the rich predictive model is trained with a training corpus for an automatic speech recognition detector and a training corpus for the natural language vocalization generator.

3. The method according to claim 1, wherein the natural language vocalization generator uses the partially completed word sequence and the predicted remainder portion for information retrieval from a non-transitory memory.

4. The method according to claim 1, wherein the rich predictive model is an n-gram model, a recurrent neural network model, a long short term memory, feed forward neural network, or a combination thereof.

5. The method according to claim 1, further comprising receiving an electrical signal at an automatic speech recognition detector, wherein the electric signal is received in real time from the audio sensor receiving the utterance.

6. The method according to claim 5, further comprising recognizing one or more words for construction of the partially completed word sequence from the received electrical signal concurrently with reception of the electrical signal.

7. The method according to claim 6, further comprising conveying the vocalization to a speaker in response to the received complete utterance.

8. A non-transitory computer readable storage medium comprising a computer readable program for reducing response time in a speech interface, wherein the computer readable program when executed on a computer causes the computer to perform the steps of:

constructing a partially completed word sequence from a partially received utterance from a speaker received by an audio sensor;

modeling a remainder portion for the partially received utterance using a processor based on a rich predictive model to predict the remainder portion; and

responding to the partially completed word sequence and the predicted remainder portion for the partially received utterance using a natural language vocalization generator with a vocalization, wherein the vocalization is prepared before a complete utterance is received from the speaker and conveyed to the speaker by an audio transducer.

9. The computer readable storage medium of claim 8, wherein the computer readable program when executed on a computer causes the computer to:

train with a training corpus for an automatic speech recognition detector and a training corpus for the natural language vocalization generator.

10. The computer readable storage medium of claim 8, wherein the computer readable program when executed on a computer causes the computer to:

use an n-gram model, a recurrent neural network model, a long short term memory, feed forward neural network, or a cornbination thereof for the rich predictive model.

11. The computer readable storage medium of claim 8, wherein the computer readable program when executed on a computer causes the computer to:

receive an electrical signal at an automatic speech recognition detector, wherein the electric signal is received in real time from the audio sensor receiving the utterance.

12. A system for reducing response time in a speech interface comprising:

an automatic speech recognition detector including a processor configured to construct a partially completed word sequence from a partially received utterance from a speaker received by an audio sensor;

a word predictor configured to model a remainder portion for the partially received utterance using a processor based on a rich predictive model to predict the remainder portion; and

a natural language vocalization generator configured to respond to the partially completed word sequence and the predicted remainder portion for the partially received utterance using a natural language vocalization generator with a vocalization, wherein the vocalization is prepared before a complete utterance is received from the speaker and conveyed to the speaker by an audio transducer.

13. The system according to claim 12, wherein the rich predictive model is trained with a training corpus for an automatic speech recognition detector and a training corpus for the natural guage utterance generator.

14. The system according to claim 12, wherein the natural language vocalization generator uses the partially completed word sequence and the predicted remainder portion for information retrieval.

15. The system according to claim 12, wherein the language model is an n-gram model, a neural network, a long short term memory, feed forward neural network, or a combination thereof stored in a non-transitory memory.

16. The system according to claim 12, wherein e automatic speech recognition detector is coupled to the audio sensor to receive an electrical signal, Wherein the electric signal is generated by the utterance from the speaker, and the processor of the automatic speech recognition detector is configured to recognize one or more words from the received signal.

17. The system according to claim 16, wherein the automatic speech recognition detector is configured to recognize one or more words for construction of the partially completed word sequence concurrently with reception of the electrical signal.

18. The system according to claim 17, wherein the natural language vocalization generator is configured to convey the vocalization to the speaker in response to the received complete utterance.

19. The system according to claim 18, wherein the system for reducing response time in speech interfacing further comprises non-transitory memory, and the vocalization to the speaker includes information retrieved from the non-transitory memory.

20. The system according to claim 19, wherein the information includes items associated with the particular speech interface use.