Alternate natural language input generation

Abstract

Techniques for handling errors during processing of natural language inputs are described. A system may process a natural language input to generate an ASR hypothesis or NLU hypothesis. The system may use more than one data searching technique (e.g., deep neural network searching, convolutional neural network searching, etc.) to generate an alternate ASR hypothesis or NLU hypothesis, depending on the type of hypothesis input for alternate hypothesis processing.

1 .- 20 . (canceled) 
     
     
         21 . A computer-implemented method, comprising:
 receiving audio data representing a spoken natural language input;   performing speech processing to determine a first speech processing hypothesis corresponding to a first interpretation of the spoken natural language input;   determining a likelihood that further processing of the first speech processing hypothesis will result in a processing error;   based at least in part on the likelihood, determining a second speech processing hypothesis corresponding to a second interpretation of the spoken natural language input different from the first interpretation;   determining the second speech processing hypothesis corresponds to an action; and   causing performance of the action.   
     
     
         22 . The computer-implemented method of  claim 21 , further comprising:
 determining the first speech processing hypothesis corresponding to a first entity; and   determining the second speech processing hypothesis corresponding to a second entity different from the first entity.   
     
     
         23 . The computer-implemented method of  claim 21 , wherein:
 performing speech processing comprises performing automatic speech recognition (ASR) using the audio data to determine a first ASR hypothesis, wherein the first speech processing hypothesis comprises the first ASR hypothesis; and   determining the second speech processing hypothesis comprises determining a second ASR hypothesis different from the first ASR hypothesis.   
     
     
         24 . The computer-implemented method of  claim 21 , wherein:
 performing speech processing comprises performing natural language understanding (NLU) using first data representing the audio data to determine a first NLU hypothesis, wherein the first speech processing hypothesis comprises the first NLU hypothesis; and   determining the second speech processing hypothesis comprises determining a second NLU hypothesis different from the first NLU hypothesis.   
     
     
         25 . The computer-implemented method of  claim 21 , further comprising:
 sending, from a speech processing component, to a first component, first data corresponding to the first speech processing hypothesis; and   processing the first data using the first component to determine the likelihood.   
     
     
         26 . The computer-implemented method of  claim 25 , further comprising:
 processing, by the first component, the first data with respect to stored data to determine the likelihood, the stored data corresponding to at least one prior speech processing hypothesis.   
     
     
         27 . The computer-implemented method of  claim 21 , further comprising:
 processing first data corresponding to the first speech processing hypothesis with respect to stored data to determine a prior speech processing hypothesis that resulted in a system performing a correct action; and   selecting the prior speech processing hypothesis as the second speech processing hypothesis.   
     
     
         28 . The computer-implemented method of  claim 27 , further comprising:
 determining a user profile corresponding to the spoken natural language input; and   determining the stored data based at least in part on the user profile.   
     
     
         29 . The computer-implemented method of  claim 21 , further comprising:
 determining that a confidence value associated with the first speech processing hypothesis fails to satisfy a confidence threshold.   
     
     
         30 . The computer-implemented method of  claim 29 , wherein determination of the second speech processing hypothesis occurs after determining that the confidence value associated with the first speech processing hypothesis fails to satisfy the confidence threshold. 
     
     
         31 . A system comprising:
 at least one processor; and   at least one memory comprising instructions that, when executed by the at least one processor, cause the system to:
 receive audio data representing a spoken natural language input; 
 perform speech processing to determine a first speech processing hypothesis corresponding to a first interpretation of the spoken natural language input; 
 determine a likelihood that further processing of the first speech processing hypothesis will result in a processing error; 
 based at least in part on the likelihood, determine a second speech processing hypothesis corresponding to a second interpretation of the spoken natural language input different from the first interpretation; 
 determine the second speech processing hypothesis corresponds to an action; and 
 cause performance of the action. 
   
     
     
         32 . The system of  claim 31 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
 determine the first speech processing hypothesis corresponding to a first entity; and   determine the second speech processing hypothesis corresponding to a second entity different from the first entity.   
     
     
         33 . The system of  claim 31 , wherein:
 the instructions that cause the system to perform speech processing comprise instructions that, when executed by the at least one processor, cause the system to perform automatic speech recognition (ASR) using the audio data to determine a first ASR hypothesis, wherein the first speech processing hypothesis comprises the first ASR hypothesis; and   the instructions that cause the system to determine the second speech processing hypothesis comprise instructions that, when executed by the at least one processor, cause the system to determine a second ASR hypothesis different from the first ASR hypothesis.   
     
     
         34 . The system of  claim 31 , wherein:
 the instructions that cause the system to perform speech processing comprise instructions that, when executed by the at least one processor, cause the system to perform natural language understanding (NLU) using first data representing the audio data to determine a first NLU hypothesis, wherein the first speech processing hypothesis comprises the first NLU hypothesis; and   the instructions that cause the system to determine the second speech processing hypothesis comprise instructions that, when executed by the at least one processor, cause the system to determine a second NLU hypothesis different from the first NLU hypothesis.   
     
     
         35 . The system of  claim 31 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
 send, from a speech processing component, to a first component, first data corresponding to the first speech processing hypothesis; and   process the first data using the first component to determine the likelihood.   
     
     
         36 . The system of  claim 35 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
 process, by the first component, the first data with respect to stored data to determine the likelihood, the stored data corresponding to at least one prior speech processing hypothesis.   
     
     
         37 . The system of  claim 31 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
 process first data corresponding to the first speech processing hypothesis with respect to stored data to determine a prior speech processing hypothesis that resulted in a system performing a correct action; and   select the prior speech processing hypothesis as the second speech processing hypothesis.   
     
     
         38 . The system of  claim 37 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
 determine a user profile corresponding to the spoken natural language input; and   determine the stored data based at least in part on the user profile.   
     
     
         39 . The system of  claim 31 , wherein the at least one memory further comprises instructions that, when executed by the at least one processor, further cause the system to:
 determine that a confidence value associated with the first speech processing hypothesis fails to satisfy a confidence threshold.   
     
     
         40 . The system of  claim 39 , wherein determination of the second speech processing hypothesis occurs after determination that the confidence value associated with the first speech processing hypothesis fails to satisfy the confidence threshold.