Methods and systems for adapting speech systems
View Patent ↗Methods and systems are provided for adapting a speech system. In one example a method includes: processing a spoken command with one or more models of one or more model types to achieve model results; evaluating a frequency of the model results; and selectively updating the one or more models of the one or more model types based on the evaluating.
1. A method of adapting a speech system, comprising:
processing a spoken command using a first speech model of a first model type to obtain a first result;
processing the spoken command using a second speech model of a second model type to obtain a second result;
evaluating a first frequency of an association of the first result and the first speech model;
evaluating a second frequency of an association of the second result and the second speech model; and
selectively updating at least one of the first speech model and the second speech model, wherein the selectively is based on the evaluation of the first frequency and the evaluation of the second frequency.
2. The method of claim 1 further comprising processing the first result and the second result using a third speech model of a third model type to obtain a third result.
3. The method of claim 2 further comprising:
selecting at least one of the first speech model and the first result, and the second speech model and the second result based on the third result; and
updating a fourth model that is a system model based on the selecting.
4. The method of claim 3 further comprising determining a third frequency of an association of the third result and the third speech model; and wherein the selecting the at least one of the first speech model and the first result, and the second speech model and the second result is based on the third frequency.
5. The method of claim 1 wherein the first model type and the second model type are different model types.
6. The method of claim 1 wherein the first speech model is a fixed language model type that recognizes a fixed list of commands.
7. The method of claim 1 wherein the second speech model is a wider scope model type that recognizes phrases.
8. The method of claim 2 wherein the third speech model is an arbitrator model type that arbitrates results from an other model.
9. The method of claim 2 wherein the first model type is a Finite State Grammar model type, the second model type is a Statistical Language Model type, and the third model type is a Structural Equation Model type.
10. The method of claim 1 wherein the selectively updating comprises selectively updating the at least one of the first speech model and the second speech model to include or exclude the spoken command based on the evaluation of the first frequency and the evaluation of the second frequency.
11. The method of claim 1 wherein the selectively updating comprises selectively updating at least one of a weight and a probability within the at least one of the first speech model and the second speech model based on the evaluation of the first frequency and the evaluation of the second frequency.
12. A system for adapting a speech system, comprising:
a non-transitory computer readable medium comprising:
a first module that, by a processor, processes a spoken command using a first speech model of a first model type to obtain a first result and that processes the spoken command using a second speech model of a second model type to obtain a second result;
a second module that, by a processor, evaluates a first frequency of an association of the first result and the first speech model, and that evaluates a second frequency of an association of the second result and the second speech model; and
a third module that, by a processor, selectively updates at least one of the first speech model and the second speech model, wherein the selectively is based on the evaluation of the first frequency and the evaluation of the second frequency.
13. The system of claim 12 wherein the first module processes the first result and the second result using a third speech model of a third model type to obtain a third result.
14. The system of claim 13 wherein the second module selects at least one of the first speech model and the first result, and the second speech model and the second result based on the third result, and wherein the third module updates a fourth model that is a system model based on the selection.
15. The system of claim 14 wherein the second module determines a third frequency of the third result, and selects the at least one of the first speech model and the first result, and the second speech model and the second result based on the third frequency.
16. The system of claim 12 wherein the first module processes the spoken command by processing the spoken command using at least three speech models of at least three different model types.
17. The system of claim 12 wherein the first speech model is a fixed language model type that recognizes a fixed list of commands.
18. The system of claim 12 wherein the second speech model is a wider scope model type that recognizes phrases.
19. The system of claim 13 wherein the third speech model is an arbitrator model type that arbitrates results from an other model.
20. The system of claim 12 wherein the third module selectively updates the at least one of the first speech model and the second speech model to include or exclude the spoken command based on the evaluation of the first frequency and the evaluation of the second frequency.
21. The system of claim 12 wherein the third module selectively updates at least one of a weight and a probability of the at least one of the first speech model and the second speech model based on the evaluation of the first frequency and the evaluation of the second frequency.