Automatic accuracy estimation for audio transcriptions

Patent metadata

Field	Value
Publication number	US-9390707-B2
Application number	US-201614997728-A
Country	US
Kind code	B2
Filing date	Jan 18, 2016
Priority date	May 3, 2012
Publication date	Jul 12, 2016
Grant date	Jul 12, 2016

Abstract

Embodiments of the present invention provide an approach for estimating the accuracy of a transcription of a voice recording. Specifically, in a typical embodiment, each word of a transcription of a voice recording is checked against a customer-specific dictionary and/or a common language dictionary. The number of words not found in either dictionary is determined. An accuracy number for the transcription is calculated from the number of said words not found and the total number of words in the transcription.

First claim

What is claimed is:

1. A computer-implemented method for estimating the accuracy of a transcription of a voice recording, comprising:
calculating an accuracy number by dividing a number of accurate words in the transcript on by a total number of words in the transcription;
assigning a greater weight to at least one word exceeding a predefined number of characters as compared to at least one other word below the predetermined number of characters;
automatically extracting using natural language processing, from a data structure that is one of a common language dictionary, customer specific dictionary or a knowledge base, a set of axioms associated with at least one of the words in the transcription, wherein the set of axioms comprises a computer-parsable definition of a relationship of data to the at least one of the words; and
assigning a confidence level to each axiom of the set of axioms based on a result of the dividing and the assigning of weight, wherein the confidence level is assigned based on an output of a Gaussian function applied to a result of the dividing and the assigning of weight.

2. The computer-implemented method of claim 1 , further comprising determining a number of inaccurate words by identifying the number of words not found in either a customer specific dictionary or a dictionary of common language words.

3. The computer-implemented method of claim 1 , further comprising determining a number of inaccurate words by indentifying the number of words determined to be spelled incorrectly.

4. The method of claim 1 , further comprising assigning a greater weight to at least one word exceeding a predefined number of characters as compared to at least one other word below the predetermined number of characters.

5. The method of claim 1 , further comprising retrieving, from a data structure, a set of axioms associated with at least one of the words in the transcription, wherein the set of axioms comprises a computer-parsable definition of a relationship of data to the at least one of the words.

6. The method of claim 4 , further comprising assigning a confidence level to each axiom of the set of axioms.

7. The method of claim 1 , wherein a solution service provider provides a computer infrasture operable to perform the method.

8. A system for estimating the accuracy of a transcription of a voice recording, comprising:
a memory medium comprising instructions; a bus coupled to the memory medium; and
an audio transcription tool coupled to the bus that when executing the instructions causes the system to:
calculate an accuracy number by dividing a number of accurate words in the transcription by a total number of words in the transcription;
assign a greater weight to at least one word exceeding a predefined number of characters as compared to at least one other word below the predetermined number of characters;
automatically extracting using natural language processing, from a data structure that is one of a common language dictionary, customer specific dictionary or a knowledge base, a set of axioms associated with at least one of the words in the transcription, wherein the set of axioms comprises a computer-parsable definition of a relationship of data to the at least one of the words; and
assign a confidence level to each axiom of the set of axioms based on a result of the dividing and the assigning of weight, wherein the confidence level is assigned based on an output of a Gaussian function applied to a result of the dividing and the assigning of the weight.

9. The system of claim 8 , the memory medium further comprising instructions for causing the system to determine a number of inaccurate words by at least one of:
identifying the number of words not found in either a customer specific dictionary or a dictionary of common language words; or
determining a number of inaccurate words by indentifying the number of words determined to be spelled incorrectly.

10. The system of claim 8 , the memory medium further comprising instructions for causing the system to assign a greater weight to at least one word exceeding a predefined number of characters as compared to at least one other word below the predetermined number of characters.

11. The system of claim 8 , the memory medium further comprising instructions for causing the system to retrieve, from a data structure, a set of axioms associated with at least one of the words in the transcription, wherein the set of axioms comprises a computer-parsable definition of a relationship of data to the at least one of the word.

12. The system of claim 10 , the memory medium further comprising instructions for causing the system to assign a confidence level to each axiom.

13. A computer program product for estimating the accuracy of a transcription of a voice recording, the computer program product comprising a non-transitory computer readable storage medium, and program instructions stored on the non-transitory computer readable storage medium, to:
calculate an accuracy number by dividing a number of accurate words in the transcription by a total number of words in the transcription;
assign a greater weight to at least one word exceeding a predefined number of characters as compared to at least one other word below the predetermined number of characters;
automatically extracting using natural language processing, from a data structure that is one of a common language dictionary, customer specific dictionary or a knowledge base, a set of axioms associated with at least one of the words in the transcription, wherein the set of axioms comprises a computer-parsable definition of a relationship of data to the at least one of the words; and
assign a confidence level to each axiom of the set of axioms based on a result of the dividing and the assigning of weight, wherein the confidence level is assigned based on an output of a Gaussian function applied to a result of the dividing and the assigning of weight.

14. The computer program product of claim 13 , the computer readable storage medium further comprising instructions to determine a number of inaccurate words by at least one of:
identifying the number of words not found in either a customer specific dictionary or a dictionary of common language words; or
determining a number of inaccurate words by indentifying the number of words determined to be spelled incorrectly.

15. The computer program product of claim 13 , the computer readable storage medium further comprising instructions to assign a greater weight to at least one word exceeding a predefined number of characters as compared to at least one other word below the predetermined number of characters.

16. The computer program product of claim 13 , the computer readable storage medium further comprising instructions to retrieve, from a data structure, a set of axioms associated with at least one of the words in the transcription, wherein the set of axioms comprises a computer-parsable definition of a relationship of data to the at least one of the words.

17. The computer program product of claim 15 , the computer readable storage medium further comprising instructions to assign a confidence level to each axiom of the set of axioms.

Assignees

• IBM

Inventors

• Bostick James E
• Ganci Jr John M
• Kaemmerer John P
• Trim Craig M

Classifications

• G10L15/01 Primary

  Assessment or evaluation of speech recognition systems · CPC title

• G10L15/26

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

• G10L15/1822

  Parsing for meaning understanding · CPC title

• G06F40/242

  Dictionaries · CPC title

• G06F40/51

  Translation evaluation · CPC title