Voice Assistant Candidate Validation for ASR Error Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing voice assistant systems fail to accurately process user inputs due to errors in Automatic Speech Recognition (ASR) and Natural Language Understanding (NLU), leading to unpredictable and undesirable outcomes, especially in noisy environments or with user-specific data, without considering overall system response status.
Innovation Solution
A method and system that uses trained AI models to identify candidate text outputs, validate their feasibility, generate alternative hypotheses based on contextual and user-specific data, and execute the most accurate output through phonetic matching and decoding, addressing failures in ASR and NLU.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ASR and NLP systems process voice input independently, then processing speed is maintained, but accuracy deteriorates due to inability to correct ASR errors
Solution Approach 1:
The patent implements a feedback mechanism where the execution validator checks the output of the action planner and provides correction feedback to the NLP system. This allows the system to detect when ASR errors lead to incorrect action execution and feed this information back for correction, improving overall accuracy without requiring complete system redesign
Solution Approach 2:
The system is divided into independent modules (ASR, NLP, Action Planner, Execution Validator) that can process voice input sequentially. Each module handles a specific function, allowing the system to maintain processing speed while enabling error correction through the validator's feedback mechanism
2Measurement precision
If NLP system performs error correction, then accuracy improves, but processing time increases
Solution Approach 1:
The system performs preliminary action by having the execution validator check for errors before the final action execution. This preliminary validation step identifies ASR errors that would lead to incorrect actions, allowing correction to be made efficiently without requiring continuous processing time extension
Solution Approach 2:
The execution validator provides feedback to the NLP system about ASR errors, enabling efficient error correction. This feedback mechanism allows the system to correct errors without requiring complete reprocessing of the entire voice input, thus minimizing time loss
3Productivity
If system executes actions based on ASR output, then productivity is maintained, but reliability deteriorates due to ASR errors
Solution Approach 1:
The execution validator checks the action planner's output against the original voice input and provides feedback about potential errors. This feedback mechanism ensures that actions are executed reliably by catching ASR errors before they lead to incorrect actions, while maintaining productivity through efficient error detection and correction
Solution Approach 2:
The system performs self-service by having the execution validator automatically detect and correct its own errors in action execution. This self-validation mechanism ensures reliable action execution without requiring external intervention, maintaining productivity while improving reliability
Data Source
AI summary
The disclosure relates to a method and apparatus for processing a voice input. The method includes identifying a first candidate text output from a first list of candidate text outputs, the first list of candidate text outputs being generated based on a voice input being received from a user. The method includes validating a feasibility of executing the first candidate text output based on an analysis of a first contextual data related to the first candidate text output, user-specific data, historical data related to previous validations of the voice assistant device, and one of an execution result or a no-action result corresponding to the first candidate text output. The method includes, based on the feasibility of executing the first candidate text output being validated as not feasible for execution, identifying failure information associated with the first candidate text output.