System and method for validating natural language content using crowdsourced validation jobs

• US 9,922,653 B2
• Filed: 07/25/2016
• Issued: 03/20/2018
• Est. Priority Date: 09/07/2015
• Status: Expired due to Fees

• Current Assignee: Cerence Inc.
• Original Assignee: VoiceBox Technologies Corporation (Microsoft Corporation)
• Inventors: Rothwell, Spencer John, Braga, Daniela, Elshenawy, Ahmad Khamis, Carter, Stephen Steele
• Primary Examiner(s): MCFADDEN, SUSAN IRIS
• Application Number:
  US15/219,088
• Publication Number:
  US 20170069326A1
• Time in Patent Office: 603 Days
• Field of Search: 704235
• CPC Class Codes: G06F 40/226 Validation
• CPC Class Codes: G06F 40/40 Processing or translation of natural language natural language analysis G06F40/20; semantic analysis G06F40/30
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 50/01 Social networking
• CPC Class Codes: G10L 15/01 Assessment or evaluation of speech recognition systems
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications

Abstract

Systems and methods of validating transcriptions of natural language content using crowdsourced validation jobs are provided herein. In various implementations, a transcription pair comprising natural language content and text corresponding to a transcription of the natural language content may be gathered. A group of validation devices may be selected for reviewing the transcription pair. A crowdsourced validation job may be created for the group of validation devices. The crowdsourced validation job may be provided to the group of validation devices. One or more votes representing whether or not the text accurately represents the natural language content may be received from the group of validation devices. Based on the one or more votes received, the transcription pair may be stored in a validated transcription library, which may be used to process end-user voice data.

Claims

1. A computer-implemented method of generating a transcription library comprising validated transcription pairs, wherein the transcription library is used to perform natural language processing, the method being implemented in a computer system having one or more physical processors programmed with computer program instructions that, when executed by the one or more physical processors, program the computer system to perform the method, the method comprising:

obtaining, by the computer system, a transcription pair comprising natural language content and text, wherein the natural language content comprises audio content received from one or more audio input components, and wherein the text corresponds to a transcription of the natural language content;

creating, by the computer system, a first crowdsourced validation job to be performed at one or more first validation devices, the first crowdsourced validation job comprising first instructions for a crowd user to provide an indication of an accuracy of the transcription of the natural language content;

causing, by the computer system, the transcription pair and the first crowdsourced validation job to be provided to the one or more first validation devices;

receiving, by the computer system, at least a first response from at least a first validation device from among the one or more first validation devices, wherein the first response includes a first indication of an accuracy of the transcription of the natural language content;

storing, by the computer system, the transcription pair in a validated transcription library based, at least in part, on the first response;

receiving, by the computer system, end-user voice data comprising a natural language utterance; and

identifying, by the computer system, one or more words of the natural language utterance based on the validated transcription library.

2. The method of claim 1, further comprising:

assigning, by the computer system, a validation score to the transcription pair based, at least in part, on the first response received from the first validation device with respect to the transcription pair, wherein the transcription pair is stored in the validated transcription library based on the validation score.

3. The method of claim 1, further comprising:

determining, by the computer system, whether or not the one or more first validation devices agreed the text is an accurate transcription of the natural language content; and

determining, by the computer system, whether to provide the transcription pair to one or more second validation devices.

4. The method of claim 3, wherein if the one or more first validation devices did not agree the text is an accurate transcription of the natural language content, the method further comprises:

causing, by the computer system, the transcription pair to be provided to the one or more second validation devices;

receiving, by the computer system, at least a second response from at least a second validation device from among the one or more second validation devices, wherein the second response includes a second indication of an accuracy of the transcription of the natural language content; and

storing, by the computer system, the transcription pair in the validated transcription library based, at least in part, on the second response.

5. The method of claim 2, further comprising:

identifying, by the computer system, a confidence score of each of the one or more first validation devices, the confidence score representing confidence in the response from the each of the one or more first validation devices; and

updating, by the computer system, the validation score with respect to the transcription pair using the confidence score.

6. The method of claim 1, wherein the one or more first validation devices is two validation devices.

7. The method of claim 1, wherein the first instructions configure the one or more first validation devices to provide the natural language content and the text in a survey on a mobile application on the one or more first validation devices.

8. The method of claim 7, wherein the survey includes one or more tasks configured to assist engagement of first validators associated with the one or more first validation devices when responding.

9. The method of claim 7, wherein the survey includes one or more of:

test questions, captions that are not machine-readable, and audio that is not understandable to machines.

10. The method of claim 1, further comprising:

identifying, by the computer system, a plurality of transcription devices performing crowdsourced transcription jobs; and

selecting, by the computer system, the one or more first validation devices from the plurality of transcription devices.

11. The method of claim 2, further comprising storing the transcription pair in the validated transcription library if the validation score of the transcription pair exceeds a validation threshold.

12. The method of claim 1, wherein the validated transcription library is used to identify the one or more words of the natural language utterance in real-time.

13. The method of claim 1, wherein causing the first crowdsourced validation job to be provided to the one or more first validation devices comprises:

causing, by the computer system, the first crowdsourced validation job to be provided to the one or more first validation devices in a validator application.

14. The method of claim 1, wherein causing the first crowdsourced validation job to be provided to the one or more first validation devices comprises:

causing, by the computer system, a message that includes a direct selectable link to an interface to be provided to the one or more first validation devices, wherein the interface includes the first crowdsourced validation job.

15. The method of claim 1, wherein the end-user voice data is received from an end-user device, the method further comprising:

causing, by the computer system, a textual representation of the one or more words to be provided to the end-user device.

16. A system configured to generate a transcription library comprising validated transcription pairs, wherein the transcription library is used to perform natural language processing, the system comprising:

one or more physical processors programmed with one or more computer program instructions which, when executed, program the one or more physical processors to;

obtain a transcription pair comprising natural language content and text, wherein the natural language content comprises audio content received from one or more audio input components, and wherein the text corresponds to a transcription of the natural language content;

create a first crowdsourced validation job to be performed at one or more first validation devices, the first crowdsourced validation job comprising first instructions for a crowd user to provide an indication of an accuracy of the transcription of the natural language content;

cause the transcription pair and the first crowdsourced validation job to be provided to the one or more first validation devices;

receive at least a first response from at least a first validation device from among the one or more first validation devices, wherein the first response includes a first indication of an accuracy of the transcription of the natural language content;

store the transcription pair in a validated transcription library based, at least in part, on the first response;

receive end-user voice data comprising a natural language utterance; and

identify one or more words of the natural language utterance based on the validated transcription library.

17. The system of claim 16, wherein the one or more physical processors are further programmed to:

assign a validation score to the transcription pair based, at least in part, on the first response received from the first validation device with respect to the transcription pair, wherein the transcription pair is stored in the validated transcription library based on the validation score.

18. The system of claim 16, wherein the one or more physical processors are further programmed to:

determine whether or not the one or more first validation devices agreed the text is an accurate transcription of the natural language content; and

determine whether to provide the transcription pair to one or more second validation devices.

19. The system of claim 18, wherein if the one or more first validation devices did not agree the text is an accurate transcription of the natural language content, the one or more processors are further programmed to:

cause the transcription pair to be provided to the one or more second validation devices;

receive at least a second response from at least a second validation device from among the one or more second validation devices, wherein the second response includes a second indication of an accuracy of the transcription of the natural language content; and

store the transcription pair in the validated transcription library based, at least in part, on the second response.

20. The system of claim 17, wherein the one or more physical processors are further programmed to:

identify a confidence score of each of the one or more first validation devices, the confidence score representing confidence in the response from the each of the one or more first validation devices; and

update the validation score with respect to the transcription pair using the confidence score.

21. The system of claim 16, wherein the one or more first validation devices is two validation devices.

22. The system of claim 16, wherein the first instructions configure the one or more first validation devices to provide the natural language content and the text in a survey on a mobile application on the one or more first validation devices.

23. The system of claim 22, wherein the survey includes one or more tasks configured to assist engagement of first validators associated with the one or more first validation devices when responding.

24. The system of claim 22, wherein the survey includes one or more of:

test questions, captions that are not machine-readable, and audio that is not understandable to machines.

25. The system of claim 16, wherein the one or more physical processors are further programmed to:

identify a plurality of transcription devices performing crowdsourced transcription jobs; and

select the one or more first validation devices from the plurality of transcription devices.

26. The system of claim 17, further comprising storing the transcription pair in the validated transcription library if the validation score of the transcription pair exceeds a validation threshold.

27. The system of claim 16, wherein the validated transcription library is used to identify the one or more words of the natural language utterance in real-time.

28. The system of claim 16, wherein to cause the first crowdsourced validation job to be provided to the one or more first validation devices, the one or more physical processors are further programmed to:

cause the first crowdsourced validation job to be provided to the one or more first validation devices in a validator application.

29. The system of claim 16, wherein to cause the first crowdsourced validation job to be provided to the one or more first validation devices, the one or more physical processors are further programmed to:

cause a message that includes a direct selectable link to an interface to be provided to the one or more first validation devices, wherein the interface includes the first crowdsourced validation job.

30. The system of claim 16, wherein the end-user voice data is received from an end-user device, wherein the one or more physical processors are further programmed to:

cause a textual representation of the one or more words to be provided to the end-user device.