Method and apparatus for improving the transcription accuracy of speech recognition software

• US 20070156403A1
• Filed: 08/25/2006
• Published: 07/05/2007
• Est. Priority Date: 03/01/2003
• Status: Active Grant

• Current Assignee: Robert Coifman
• Original Assignee: Robert Coifman
• Inventors: Coifman, Robert
• Granted Patent: US 7,809,565 B2
• Time in Patent Office: Days
• US Class Current: 704/252
• CPC Class Codes: G10L 15/08 Speech classification or search

Abstract

The present invention involves an enhanced method for operating a speech recognition system in which sequential vocabularies are loaded for comparison to the input speech. Form each sequential vocabulary a subset of candidate vocabulary elements are selected for matching, the probability match scores of each of the selected candidate vocabulary elements is weighted by a weighting factor that is dependent on the particular vocabulary from which the vocabulary elements are selected. As each set of candidate vocabulary elements is selected from the next sequential vocabulary, the selected set is combined with the set of previously selected and weighted vocabulary elements, which is then further weighted according to the weighting criteria of the then active vocabulary. At the end of the sequential vocabulary selection grouping process, the final, sequentially weighted match scores for the candidate vocabulary elements are arranged and an appropriate match to the input speech is presented.

Claims

1. A method for operating a computerized speech recognition system comprising:

loading a first vocabulary;

evaluating individual vocabulary elements within said first vocabulary to determine a first vocabulary match set, each vocabulary element within said first vocabulary match set having a match probability score;

weighting said match probability scores of said vocabulary elements within said first vocabulary match set with a first vocabulary weighting factor;

loading a second vocabulary;

evaluating individual vocabulary elements within said second vocabulary to determine a second vocabulary match set, each vocabulary element within said second vocabulary match set having a match probability score;

combining said individual vocabulary elements within said first and second vocabulary match sets so as to create a combine set of vocabulary elements;

weighting said match probability scores of said combine set of vocabulary elements with a second vocabulary weighting factor; and

selecting as a match to an input to said computerized speech recognition system a vocabulary element from said combine set of vocabulary elements based on said weighted match probability scores of said combine set of vocabulary elements.

2. The method of claim 1 further comprising:

reducing a size of said combine set of vocabulary elements to create a reduced combine set of vocabulary elements.

3. The method of claim 1 further comprising:

loading a third vocabulary;

evaluating individual vocabulary elements within said third vocabulary to determine a third vocabulary match set, each vocabulary element within said third vocabulary match set having a match probability score; and

combining said individual vocabulary elements with said combined set of vocabulary elements so as to create a new combine set of vocabulary elements.

4. The method of claim 1 wherein said first and second weighting functions are linear scaling factors and said step of weighting includes the step of multiplying said match probability score by said linear scaling factors.

5. The method of claim 1 wherein said first and second weighting functions are non-linear scaling factors and said step of weighting includes the step of applying the non-linear scaling factor to said match probability score.

6. The method of claim 1 wherein said first and second vocabularies are selected based on a user of said speech recognition system.

7. The method of claim 1 wherein said first and second vocabularies are selected according to a particular form field being populated by said speech recognition system.

8. The method of claim 1 wherein said first and second vocabularies are selected according to a speech context being used by a user of said speech recognition system.

10. The method of claim 7 wherein said step of weighting further includes:

applying said non-linear weighting function to said match probability scores of said vocabulary elements within said first and second vocabulary match sets;

calculating a first altered match probability score for said vocabulary elements within said first vocabulary match set;

deriving a second altered match probability score for said vocabulary elements within said second vocabulary match set; and

deriving modified first and second match probability scores for said vocabulary elements within said combined set of vocabulary elements.

11. The method of claim 8 wherein said first and second vocabularies are selected according to a user of said speech recognition system.

12. The method of claim 8 wherein said first and second vocabularies are selected according to a particular form field being populated by said speech recognition system.

13. The method of claim 8 wherein said first and second vocabularies are selected according to a speech context being used by a user of said speech recognition system.

9. A method for operating a computerized speech recognition system comprising:

loading a first vocabulary;

evaluating individual vocabulary elements within said first vocabulary to determine a first vocabulary match set, each vocabulary element within said first vocabulary match set having a match probability score;

loading a second vocabulary;

evaluating individual vocabulary elements within said second vocabulary to determine a second vocabulary match set, each vocabulary element within said second vocabulary match set having a match probability score;

combining said individual vocabulary elements within said first and second vocabulary match sets so as to create a combine set of vocabulary elements;

weighting said match probability scores of said combine set of vocabulary elements with a non-linear vocabulary weighting function;

evaluating individual vocabulary elements within said combined set of vocabulary elements to determine a combined vocabulary match set based on said non-linearly weighted match probability scores of said vocabulary element within combined set of vocabulary elements; and

selecting as a match to an input to said computerized speech recognition system a vocabulary element from said combine set of vocabulary elements based on said weighted match probability scores of said combine set of vocabulary elements.