US6178401B1
Expired / Lapsed
US
⏱️ 0.0 years remaining
Method for reducing search complexity in a speech recognition system
Assignee: Nuance Communications Inc
•
Filing: 1998-08-28
•
Granted: 2001-01-23
•
Expires: 2018-08-28
Abstract
Abstract A method is provided for reducing search complexity in a speech recognition system having a fast match, a detailed match, and a language model. Based on at least one predetermined variable, the fast match is optionally employed to generate candidate words and acoustic scores corresponding to the candidate words. The language model is employed to generate language model scores. The acoustic scores are combined with the language model scores and the combined scores are ranked to determine top ranking candidate words to be later processed by the detailed match, when the fast match is employed. The detailed match is employed to generate detailed match scores for the top ranking candidate words.
Inventors
- 👤 Martin Franz
- 👤 Miroslav Novak
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-09 11:35
CPC Classifications
G; G10; G10L; G10L15/00; G10L15/08; G10L15/18; G10L15/183; G10L2015/085
Claims Analysis
Total Claims: 24
Claim 1
Independent Claim
113 words
A method for reducing search complexity in a speech recognition system having a fast match, a detailed match, and a language model, comprising the steps of: (a) optionally employing the fast match to generate candidate words and acoustic scores corresponding to the candidate words, based on at least one predetermined variable; (b) employing the language model to generate language model scores; (c) combining the acoustic scores with the language model scores and ranking the combined scores to determine top ranking candidate words to be later processed by the detailed match, when the fast match is employed; and (d) employing the detailed match to generate detailed match scores for the top ranking candidate words.
Claim 2
Independent Claim
113 words
A method for reducing search complexity in a speech recognition system having a fast match, a detailed match, and a language model, comprising the steps of: (a) optionally employing the fast match to generate candidate words and acoustic scores corresponding to the candidate words, based on at least one predetermined variable; (b) employing the language model to generate language model scores; (c) combining the acoustic scores with the language model scores and ranking the combined scores to determine top ranking candidate words to be later processed by the detailed match, when the fast match is employed; and (d) employing the detailed match to generate detailed match scores for the top ranking candidate words.
Claim 3
Dependent Claim
37 words
The method of claim 1 , wherein the language model scores generated at step (b) correspond to the candidate words generated by the fast match at step (a), when the fast match is employed at step (a).
Claim 4
Dependent Claim
70 words
The method of claim 1 , wherein said step (b) further comprises the step of generating candidate words corresponding to the language model scores, when the fast match is not employed, and the method further comprises the step of, before said step (d), ranking the language model scores to determine the top ranking candidate words to be later processed by the detailed match, when the fast match is not employed.
Claim 5
Dependent Claim
100 words
The method of claim 1 , further comprising the steps of: (e) after said step (b) and before said step (c), optionally employing the fast match to generate the candidate words and the acoustic scores corresponding to the candidate words, based on the at least one predetermined variable and the language model scores generated at step (b); and (f) combining the acoustic scores of step (e) with the language model scores and ranking the combined scores to determine the top ranking candidate words to be later processed by the detailed match, when the fast match is employed at step (e).
Claim 6
Dependent Claim
49 words
The method of claim 4 , further comprising the step of: (g) after said step (e) and before said step (f), optionally employing the language model to generate language model scores, based on the language model scores generated at step (b) and the acoustic scores generated at step (e).
Claim 7
Dependent Claim
53 words
The method of claim 5 , wherein said combining step (f) combines the language model scores generated at step (g) with the acoustic scores generated at step (e) to determine the top ranking candidate words to be later processed by the detailed match, when the language model is employed at said step (g).
Claim 8
Dependent Claim
37 words
The method of claim 5 , wherein the language model is optionally employed at said step (g) based on a model reflecting a shape of a curve of the language model scores generated at said step (b).
Claim 9
Dependent Claim
34 words
The method of claim 5 , wherein the language model is optionally employed at said step (g) based on whether the language model scores generated at said step (b) are above a predefined threshold.
Claim 10
Dependent Claim
48 words
The method of claim 4 , wherein the fast match is employed at said step (e), when the at least one predetermined variable indicates that a word context does not contain sufficient information to provide the candidate words and the language model scores are below a predefined threshold.
Claim 11
Dependent Claim
68 words
The method of claim 1 , wherein the at least one predetermined variable is a distance from a beginning of a sentence, words in a history, the acoustic, language model and detailed match scores generated from the fast match, the language model, and the detailed match, respectively, for the words in the history, and a relation of scores of a current path with respect to other active paths.
Claim 12
Dependent Claim
27 words
The method of claim 1 , wherein the at least one predetermined variable corresponds to whether a word context contain sufficient information to provide the candidate words.
Claim 13
Independent Claim
141 words
A method for reducing search complexity in a speech recognition system having a fast match, a detailed match, and a language model, comprising the steps of: (a) optionally employing the fast match to generate candidate words and acoustic scores corresponding to the candidate words, based on at least one predetermined variable; (b) employing the language model to generate language model scores; (c) combining the acoustic scores with the language model scores and ranking the combined scores to determine top ranking candidate words to be later processed by the detailed match, when the fast match is employed; (d) ranking the language model scores to determine the top ranking candidate words to be later processed by the detailed match, when the fast match is not employed; and (e) employing the detailed match to generate detailed match scores for the top ranking candidate words.
Claim 14
Independent Claim
141 words
A method for reducing search complexity in a speech recognition system having a fast match, a detailed match, and a language model, comprising the steps of: (a) optionally employing the fast match to generate candidate words and acoustic scores corresponding to the candidate words, based on at least one predetermined variable; (b) employing the language model to generate language model scores; (c) combining the acoustic scores with the language model scores and ranking the combined scores to determine top ranking candidate words to be later processed by the detailed match, when the fast match is employed; (d) ranking the language model scores to determine the top ranking candidate words to be later processed by the detailed match, when the fast match is not employed; and (e) employing the detailed match to generate detailed match scores for the top ranking candidate words.
Claim 15
Dependent Claim
68 words
The method of claim 12 , wherein the at least one predetermined variable is a distance from a beginning of a sentence, words in a history, the acoustic, language model and detailed match scores generated from the fast match, the language model, and the detailed match, respectively, for the words in the history, and a relation of scores of a current path with respect to other active paths.
Claim 16
Dependent Claim
27 words
The method of claim 12 , wherein the at least one predetermined variable corresponds to whether a word context contain sufficient information to provide the candidate words.
Claim 17
Independent Claim
206 words
A method for reducing search complexity in a speech recognition system having a fast match, a detailed match, and a language model, comprising the steps of: (a) optionally employing the fast match to generate candidate words and acoustic scores corresponding to the candidate words, based on at least one predetermined variable; (b) employing the language model to generate language model scores; (c) optionally employing the fast match to generate the candidate words and the acoustic scores corresponding to the candidate words, based on the at least one predetermined variable and the language model scores, when the fast match is not employed at step (a); (d) combining the acoustic scores generated at step (a) with the language model scores and ranking the combined scores to determine top ranking candidate words to be later processed by the detailed match, when the fast match is employed at step (a); and (e) combining the acoustic scores generated at step (c) with the language model scores and ranking the combined scores to determine top ranking candidate words to be later processed by the detailed match, when the fast match is employed at step (c); and (f) employing the detailed match to generate detailed match scores for the top ranking candidate words.
Claim 18
Independent Claim
206 words
A method for reducing search complexity in a speech recognition system having a fast match, a detailed match, and a language model, comprising the steps of: (a) optionally employing the fast match to generate candidate words and acoustic scores corresponding to the candidate words, based on at least one predetermined variable; (b) employing the language model to generate language model scores; (c) optionally employing the fast match to generate the candidate words and the acoustic scores corresponding to the candidate words, based on the at least one predetermined variable and the language model scores, when the fast match is not employed at step (a); (d) combining the acoustic scores generated at step (a) with the language model scores and ranking the combined scores to determine top ranking candidate words to be later processed by the detailed match, when the fast match is employed at step (a); and (e) combining the acoustic scores generated at step (c) with the language model scores and ranking the combined scores to determine top ranking candidate words to be later processed by the detailed match, when the fast match is employed at step (c); and (f) employing the detailed match to generate detailed match scores for the top ranking candidate words.
Claim 19
Dependent Claim
68 words
The method of claim 15 , wherein the at least one predetermined variable is a distance from a beginning of a sentence, words in a history, the acoustic, language model and detailed match scores generated from the fast match, the language model, and the detailed match, respectively, for the words in the history, and a relation of scores of a current path with respect to other active paths.
Claim 20
Dependent Claim
27 words
The method of claim 15 , wherein the at least one predetermined variable corresponds to whether a word context contain sufficient information to provide the candidate words.
Claim 21
Dependent Claim
49 words
The method of claim 15 , further comprising the step of: (g) after said step (c) and before said step (d), optionally employing the language model to generate language model scores, based on the language model scores generated at step (b) and the acoustic scores generated at step (c).
Claim 22
Dependent Claim
53 words
The method of claim 18 , wherein said combining step (e) combines the language model scores generated at step (g) with the acoustic scores generated at step (c) to determine the top ranking candidate words to be later processed by the detailed match, when the language model is employed at said step (g).
Claim 23
Dependent Claim
37 words
The method of claim 18 , wherein the language model is optionally employed at said step (g) based on a model reflecting a shape of a curve of the language model scores generated at said step (b).
Claim 24
Dependent Claim
34 words
The method of claim 18 , wherein the language model is optionally employed at said step (g) based on whether the language model scores generated at said step (b) are above a predefined threshold.
References Cited (Backward Citations)
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 2004-01-23 | 2004-07-24 | 2005-01-23 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2008-01-23 | 2008-07-24 | 2009-01-23 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2012-01-23 | 2012-07-24 | 2013-01-23 | $7700.00 | Window Closed / Lapsed |