Creating a speech recognition grammar for alphanumeric concepts

• US 7,599,837 B2
• Filed: 09/15/2004
• Issued: 10/06/2009
• Est. Priority Date: 09/15/2004
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Cecys, Mark, Ju, Yun-Cheng, Acero, Alejandro, Wang, Ye-Yi, Collins, Leonard Alan
• Primary Examiner(s): Hudspeth; David R
• Assistant Examiner(s): Neway; Samuel G
• Application Number:
  US10/941,439
• Publication Number:
  US 20060069547A1
• Time in Patent Office: 1,847 Days
• Field of Search: None
• US Class Current: 704/257
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/193 Formal grammars, e.g. finite state automata, context free grammars or word networks

Abstract

A method and system to generate a grammar adapted for use by a speech recognizer includes receiving a representation of an alphanumeric expression. For instance, the representation can take the form of a regular expression or a mask. The grammar is generated based on the representation.

Claims

1. A computer implemented method to generate a grammar adapted for use by a speech recognizer, the method comprising:

parsing by a computer a representation of an alphanumeric expression into a plurality of branches, wherein the representation is distinct from the alphanumeric expression and specifies one or more characters and a location of the one or more characters relative to each other in the alphanumeric expression, wherein the representation is of the form of a regular expression as defined by W3C;

generating grammar rules for each of the plurality of branches; and

identifying one or more smaller portions of each of the branches and generating grammar rules for each of the smaller portions, wherein generating grammar rules for each of the smaller portions comprises generating grammar rules for at least one of alternative character spoken expressions and alternative character sequence spoken expressions for the smaller portions.

2. The computer implemented method of claim 1 wherein the generating the grammar rules for alternative character spoken expressions includes using a mechanism to provide a normalized output from a speech recognizer for recognition of alternative character spoken expressions.

3. The computer implemented method of claim 2 wherein generating the grammar includes providing normalization information in the grammar so as to provide a normalized output from a speech recognizer for recognition of alternative character spoken expressions.

4. The computer implemented method of claim 3 wherein generating the grammar includes generating a database associated with the grammar, the database having normalization information.

5. The computer implemented method of claim 1 wherein the generating the grammar rules for alternative character sequence spoken expressions includes using a mechanism to provide a normalized output from a speech recognizer for recognition of alternative character sequence spoken expressions.

6. The computer implemented method of claim 5 wherein generating the grammar includes providing normalization information in the grammar so as to provide a normalized output from a speech recognizer for recognition of alternative character sequence spoken expressions.

7. The computer implemented method of claim 6 wherein generating the grammar includes generating a database associated with the grammar, the database having normalization information.

8. The computer implemented method of claim 1 wherein generating grammar rules for each of the smaller portions includes identifying a first smaller portion that is identical to a second smaller portion in the representation that grammar rules have previously been based on and generating rules based on the first smaller portion by using the rules based on the second smaller portion.

9. The computer implemented method of claim 1 wherein generating the grammar rules includes generating prefix optimized grammar rules.

10. A computer readable storage medium having instructions operable on a computer to generate a grammar adapted for use by a speech recognizer, the instructions comprising:

parsing a representation of an alphanumeric expression into a plurality of branches, wherein the representation is distinct from the alphanumeric expression and specifies one or more characters and a location of the one or more characters relative to each other in the alphanumeric expression, wherein the representation is of the form of a regular expression as defined by W3C;

generating grammar rules for each of the plurality of branches; and

identifying one or more smaller portions of each of the branches and generating grammar rules for each of the smaller portions, wherein generating grammar rules for each of the smaller portions comprises generating grammar rules for at least one of alternative character spoken expressions and alternative character sequence spoken expressions for the smaller portions.

11. The computer readable storage medium of claim 10 wherein the representation comprises a regular expression or a mask.

15. The computer readable storage medium of claim 10 wherein the generating the grammar rules for alternative character spoken expressions includes using a mechanism to provide a normalized output from a speech recognizer for recognition of alternative character spoken expressions.

16. The computer readable storage medium of claim 15 wherein generating the grammar includes providing normalization information in the grammar so as to provide a normalized output from a speech recognizer for recognition of alternative character spoken expressions.

17. The computer readable storage medium of claim 16 wherein generating the grammar includes generating a database associated with the grammar, the database having normalization information.

18. The computer readable storage medium of claim 10 wherein the generating the grammar rules for alternative character sequence spoken expressions includes using a mechanism to provide a normalized output from a speech recognizer for recognition of alternative character sequence spoken expressions.

19. The computer readable storage medium of claim 18 wherein generating the grammar includes providing normalization information in the grammar so as to provide a normalized output from a speech recognizer for recognition of alternative character sequence spoken expressions.

20. The computer readable storage medium of claim 19 wherein generating the grammar includes generating a database associated with the grammar, the database having normalization information.

12. A computer system to generate a grammar adapted for use by a speech recognizer, the computer system comprising:

one or more processing units comprising;

a parsing module adapted to parse a representation of an alphanumeric expression into a plurality of branches, wherein the representation is distinct from the alphanumeric expression and specifies one or more characters and a location of the one or more characters relative to each other in the alphanumeric expression, wherein the representation is of a regular expression as defined by W3C; and

a branch rule generator module adapted to generate rules for the grammar based on the branches; and

a piece rule generator module adapted to identify pieces of each of the branches and generate rules for the grammar based on each of the pieces; and

one or more memory units comprising;

a library for storing grammar rules based on pieces and wherein the piece rule generator module is adapted to identify a piece of the representation corresponding to a piece having rules stored in the library, and wherein the piece rule generator module is adapted to generate rules for the grammar based on the rules in the library,

and wherein the piece rule generator module is adapted to generate rules for the grammar based on at least one of alternative character spoken expressions and alternative character sequence spoken expressions for the piece.

13. The system of claim 12 wherein the piece rule generator module is adapted to store rules in the library for a first piece of the representation that has been processed, and wherein the piece rule generator module is adapted to identify a second piece of the representation that is identical to the first piece of the representation and generate rules for the second piece of the representation based on the rules in the library for the first piece of the representation.

14. The system of claim 13 wherein the piece rule generator module is adapted to provide normalization information in the grammar for at least one of alternative character spoken expressions and alternative character sequence spoken expressions.