Background speech recognition assistant

1. A method comprising:

receiving an acoustic input signal at a first stage recognizer, the first stage recognizer being configured to recognize the acoustic input signal in an always on mode;

classifying, by a computing device, portions of the acoustic input signal into different classifications using a first speech recognition algorithm, the different classifications corresponding to different topics of interest, each classification being associated with a list of keywords for its corresponding topic of interest;

determining, based on the classifying, that a particular topic of interest is being discussed in the acoustic input signal; and

triggering, in response to the determining, a second stage recognizer to turn on in order to recognize the acoustic input signal, wherein the second stage recognizer uses a second speech recognition algorithm that is different from the first speech recognition algorithm, wherein the second stage recognizer recognizes the acoustic input signal based on the particular topic of interest, and

wherein the second stage recognizer is configured to, upon recognizing the acoustic input signal:

evaluate one or more potential responses to the recognized acoustic input signal;

if at least one response in the one or more potential responses is deemed acceptable, output the at least one response to a user that originated the acoustic input signal; and

if no response in the one or more potential responses is deemed acceptable, refrain from outputting anything to the user.

2. The method of claim 1 , wherein the first stage recognizer uses a vocabulary that is smaller than a vocabulary used by the second stage recognizer.

3. The method of claim 1 , wherein the first stage recognizer uses less memory and computer processing unit (CPU) cycles than the second stage recognizer.

4. The method of claim 1 , wherein the first stage recognizer operates in a lower power mode than the second stage recognizer.

5. The method of claim 1 wherein the classifying takes into account frequency, order, and distance in time of phrases in the acoustic input signal that match classification keywords.

6. A method comprising:

receiving, by a computing device, a signal from a first stage recognizer based on recognition of an acoustic input signal and classification of portions of the acoustic input signal into different classifications using a first speech recognition algorithm, wherein the first stage recognizer is configured to recognize the acoustic input signal in an always on mode, wherein the different classifications correspond to different topics of interest, and wherein each classification is associated with a list of keywords for its corresponding topic of interest;

activating, by the computing device, a second stage recognizer upon receiving the signal in order to recognize the acoustic input signal, wherein the second stage recognizer uses a second speech recognition algorithm that is different from the first speech recognition algorithm, and wherein the second stage recognizer recognizes the acoustic input signal based on a particular topic of interest that is determined, by the first stage recognizer, to be discussed in the acoustic input signal;

evaluating one or more potential responses to the recognized acoustic input signal;

if at least one response in the one or more potential responses is deemed acceptable, outputting the at least one response to a user that originated the acoustic input signal; and

if no response in the one or more potential responses is deemed acceptable, refraining from outputting anything to the user.

7. The method of claim 6 , wherein evaluating the one or more potential responses comprises:

ranking the one or more potential responses based on one or more criteria.

8. The method of claim 7 , wherein a response in the one or more potential responses is deemed acceptable if its ranking crosses a predetermined threshold.

9. The method of claim 6 , further comprising:

determining an output method in a plurality of output methods; and

outputting the at least one response based on the output method.

10. The method of claim 9 , wherein the output method is determined based on a relevance factor, urgency factor, and an importance factor assigned to the at least one response.

11. The method of claim 6 , wherein the acoustic input signal is received through a network from the first stage recognizer.

12. The method of claim 6 wherein the classifications are based on frequency, order, and distance in time of phrases in the acoustic input signal that match classification keywords.

13. A system comprising:

a first stage recognizer operating in an always on mode, the first stage recognizer being configured to:

receive an acoustic input signal;

classify portions of the acoustic input signal into different classifications using a first speech recognition algorithm, the different classifications corresponding to different topics of interest, each classification being associated with a list of keywords for its corresponding topic of interest;

determine, based on the classifying, that a particular topic of interest is being discussed in the acoustic input signal; and

trigger, in response to the determining, a second stage recognizer to turn on in order to recognize the acoustic input signal; and

a second stage recognizer configured to:

recognize, in response to the triggering performed by the first stage recognizer, the acoustic input signal, wherein the second stage recognizer uses a second speech recognition algorithm that is different from the first speech recognition algorithm, and wherein the second stage recognizer recognizes the acoustic input signal based on the particular topic of interest;

evaluate one or more potential responses to the recognized acoustic input signal;

if at least one response in the one or more potential responses is deemed acceptable, output the at least one response to a user that originated the acoustic input signal; and

if no response in the one or more potential responses is deemed acceptable, refrain from outputting anything to the user.

14. The system of claim 13 , wherein the first stage recognizer and the second stage recognizer are located in a same device.

15. The system of claim 14 , wherein:

the device includes a processor and memory,

the first stage recognizer uses less memory and less processing cycles of the processor during speech recognition.

16. The system of claim 13 , wherein the first stage recognizer is located in a first device and the second stage recognizer is located in a second device.

17. The system of claim 16 , wherein the first device receives the acoustic input signal and sends the acoustic input signal to the second device over a network.

18. The system of claim 13 wherein the first stage recognizer classifies the portions of the acoustic input signal based on frequency, order, and distance in time of phrases in the acoustic input signal that match classification keywords.