Dynamically adjusting speech grammar weights based on usage

• US 20070208555A1
• Filed: 03/06/2006
• Published: 09/06/2007
• Est. Priority Date: 03/06/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Vila, Roberto, Ruback, Harvey, Blass, Oscar
• Granted Patent: US 8,131,548 B2
• Time in Patent Office: Days
• US Class Current: 704/9
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• 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 2015/0631 Creating reference templates; Clustering
• CPC Class Codes: G10L 2015/228 of application context

Abstract

A speech processing method can automatically and dynamically adjust speech grammar weights at runtime based upon usage data. Each of the speech grammar weights can be associated with an available speech command contained within a speech grammar to which the speech grammar weights apply. The usage data can indicate a relative frequency with which each of the available speech commands is utilized.

Claims

1. A speech processing method comprising:

automatically and dynamically adjusting speech grammar weights at runtime based upon usage data.

2. The method of claim 1, wherein each of the speech grammar weights is associated with an available speech command contained within a speech grammar to which the speech grammar weights apply, and wherein the usage data indicates a relative frequency with which each of the available speech commands is utilized.

3. The method of claim 2, wherein the speech grammar weights are adjusted so that as frequency of use for a speech command increases relative to other ones of the speech commands in the speech grammar, a speech grammar weight associated with the speech command automatically increases.

4. The method of claim 2, wherein each of the speech commands initiates a programmatic action within a system, wherein the system includes at least one non-speech, user-selectable mechanism for initiating each of the programmatic actions, wherein a relative frequency indicated by the usage data is increased when a corresponding programmatic action is initialized by the associated non-speech mechanism as well as when initialized by the associated speech command.

5. The method of claim 1, wherein the speech grammar weights and the usage data are user-specific.

6. The method of claim 5, wherein the speech grammar weights and the usage data are application specific.

7. The method of claim 6, wherein an application for which the speech grammar weights and the usage data apply is associated with a plurality of different speech grammars, wherein the adjusting step is automatically and dynamically performed for each of the different speech grammars.

8. The method of claim 5, wherein the speech grammar for which the speech grammar weights apply is a context dependent speech grammar.

9. The method of claim 5, wherein a speech grammar to which the speech grammar weights apply is utilized by a plurality of different users, wherein each of the different users is associated with a different set of user-specific speech grammar weights adjusted based upon usage data for that user.

10. The method of claim 5, wherein the speech grammar weights are associated with a speech grammar, and wherein the speech grammar is associated with a first speech-enabled application, said method further comprising:

identifying a second speech-enabled application having a speech grammar and corresponding speech grammar weights similar to the first speech-enabled application;

sharing usage data for a common user between the first speech-enabled application and the second speech-enabled application; and

responsive to the sharing step, automatically and dynamically adjusting speech grammar weights associated with at least one of the first speech-enabled application and the second speech-enabled application.

11. The method of claim 5, further comprising:

conveying at least one of the usage data and the speech grammar weights to a software developer having responsibilities for a speech grammar to which the speech grammar weights apply.

12. The method of claim 1, wherein the usage data is system specific and user independent.

13. The method of claim 1, wherein a speech grammar to which the speech grammar weights apply is utilized by a plurality of different users;

the method further comprising;

associating each of the plurality of users with a usage category, wherein each of the different usage categories is associated with a different set of category-specific speech grammar weights adjusted based upon usage data for users in that usage category.

14. The method of claim 1, wherein said steps of claim 1 are performed by at least one machine in accordance with at least one computer program having a plurality of code sections that are executable by the at least one machine.

15. The method of claim 1, wherein the steps of claim 1 are performed by at least one of a service agent and a computing device manipulated by the service agent, the steps being performed in response to a service request.

16. A software method for performing speech recognition comprising:

identifying a user;

retrieving a set of user-specific speech grammar weights from a data store, wherein the set of speech grammar weights is automatically and dynamically adjusted at runtime based upon user-specific usage data; and

speech recognizing a user utterance using a speech grammar containing a plurality of available speech commands, wherein the speech recognizing step applies the retrieved speech grammar weights to the available speech commands.

17. The software method of claim 16, further comprising:

identifying a different user;

retrieving a different set of speech grammar weights from the data store; and

speech recognizing a user utterance using the speech grammar, wherein the speech recognizing step applies the different set of speech grammar weights to the available speech commands.

18. The method of claim 16, wherein the steps of claim 16 are performed by at least one of a service agent and a computing device manipulated by the service agent, the steps being performed in response to a service request.

19. A speech recognition system comprising:

at least one speech grammar having a plurality of elements, each element consisting of at least one of a word and a phrase;

at least one set of speech grammar weights, each speech grammar element is associated with a speech grammar weight from the set;

a set of usage statistics, wherein each usage statistic is associated with one of the speech grammar weights from the set; and

programmatic instructions stored in a computer readable medium that automatically adjust the speech grammar weights in accordance with associated ones of the usage statistics, wherein the set of usage statistics is dynamically updated as the speech recognition system is utilized, and wherein the set of speech grammar weights is dynamically updated in a corresponding fashion.

20. The system of claim 19, wherein the at least one set of speech grammar weights comprise a plurality of sets of speech grammar weights, each set having a corresponding set of usage statistics, wherein each set of usage statistics is associated with at least one of a user and a category of users.