CONCISE DYNAMIC GRAMMARS USING N-BEST SELECTION

• US 20110202343A1
• Filed: 04/28/2011
• Published: 08/18/2011
• Est. Priority Date: 06/15/1998
• Status: Active Grant

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Inventors: Brown, Deborah W., Goldberg, Randy G., Marcus, Stephen Michael, Rosinski, Richard R.
• Granted Patent: US 8,682,665 B2
• Time in Patent Office: Days
• US Class Current: 704/256.1
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/06 Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice G10L15/14 takes precedence
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/19 Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules
• CPC Class Codes: G10L 15/197 Probabilistic grammars, e.g. word n-grams
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/228 of application context
• CPC Class Codes: G10L 25/27 characterised by the analysis technique

Abstract

A method and apparatus derive a dynamic grammar composed of a subset of a plurality of data elements that are each associated with one of a plurality of reference identifiers. The present invention generates a set of selection identifiers on the basis of a user-provided first input identifier and determines which of these selection identifiers are present in a set of pre-stored reference identifiers. The present invention creates a dynamic grammar that includes those data elements that are associated with those reference identifiers that are matched to any of the selection identifiers. Based on a user-provided second identifier and on the data elements of the dynamic grammar, the present invention selects one of the reference identifiers in the dynamic grammar.

Claims

1. A method comprising:

receiving speech input from a user, wherein the speech input comprises at least one non-alphanumeric character;

creating, via a processor, a correlation table based on the speech input, the correlation table listing alternative character combinations of the speech input;

selecting an alternative character combination from the correlation table most likely to match the speech input;

comparing the alternative character combination to a listing of reference identifiers to yield a matching reference; and

extracting user data associated with the matching reference from a database.

2. The method of claim 1, further comprising identifying the user based on the user data.

3. The method of claim 1, wherein the correlation table listing the alternative character combinations of the speech input is based on at least one of similar pronunciations of alphanumeric characters and alphanumeric combinations equivalent to non-alphanumeric characters.

4. The method of claim 1, wherein the correlation table listing the alternative character combinations uses a Hidden Markov Model algorithm based on the speech input.

5. The method of claim 1, further comprising analyzing a speech recognition error probability to determine the alternative character combination from the correlation table most likely to match the speech input.

6. The method of claim 5, wherein the speech recognition error probability comprises likelihood that the speech input was incorrectly recognized by a speech recognizer.

7. The method of claim 1, wherein the reference identifiers correspond to an account number.

8. A system comprising:

a processor;

a first module configured to control the processor to receive input from a user, wherein the input comprises at least one non-alphanumeric character;

a second module configured to control the processor to create a probability matrix using alternative character combinations based on the speech input;

a third module configured to control the processor to compare the probability matrix to a listing of reference identifiers to yield a highest probability combination;

a fourth module configured to control the processor to extract user data associated with the highest probability combination from a database.

9. The system of claim 8, the speech input further comprising alphanumeric characters.

10. The system of claim 8, wherein probability matrix values rely at least in part on a probability of an input recognition error.

11. The system of claim 8, wherein the input is received from the user by a speech recognition device, a telephone keypad, a touch-screen, or a remote point of sale terminal.

12. The system of claim 8, further comprising:

a fifth module configure to identify the user based at least in part on the user data.

13. The system of claim 8, wherein the second module uses a Hidden Markov Model algorithm in creating the probability matrix.

14. The system of claim 8, wherein the reference identifiers correspond to an account number.

15. A non-transitory computer-readable medium storing a computer program having instructions which, when executed by a processor, control the processor to extract user data, the instructions comprising:

receiving a speech input from a user, wherein the speech input comprises at least one non-alphanumeric character;

creating a correlation table based on the speech input, wherein the correlation table lists alternative character combinations of the speech input;

selecting an alternative character combination from the correlation table most likely to match the speech input;

comparing the alternative character combination to a listing of reference identifiers to yield a matching reference; and

extracting user data from a database, where the user data is associated with the matching reference.

16. The non-transitory computer-readable medium of claim 15, the instructions further comprising:

identifying the user based on the user data.

17. The non-transitory computer-readable medium of claim 15, wherein the alternative character combinations listed in the correlation table are based on at least one of similar pronunciations and alphanumeric combinations equivalent to non-alphanumeric characters.

18. The non-transitory computer-readable medium of claim 15, wherein the at least one non-alphanumeric character of the speech input is one of “

#”

, “

$”

, “

@”

, “

%”

, “

˜

”

, and “

_”

.

19. The non-transitory computer-readable medium of claim 15, further comprising the use of a Hidden Markov Model in creating the correlation table.

20. The non-transitory computer-readable medium of claim 15, wherein prior to selecting an alternative combination most likely to match the speech input, the processor analyzes the likelihood of speech recognition error.