Speech Recognition Correction Mechanism for Intent Accuracy

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current speech recognition systems face challenges in accurately identifying user intents from utterances and providing appropriate content services, leading to suboptimal user experience due to limitations in interpreting and correcting operations based on user inputs.

Innovation Solution

An electronic device equipped with a microphone and a processor that analyzes user utterances by obtaining commands, executing the command with the highest confidence score, and identifying subsequent utterances for correction, allowing for dynamic operation adjustments based on user feedback.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If the system executes the command with the highest confidence score without correction, then the operation speed is improved, but the accuracy of user intent interpretation deteriorates

Engineering Contradiction:
Improveoperation speedVSAvoiduser intent interpretation accuracy
Core Design Contradiction:
SpeedVSMeasurement precision

Solution Approach 1:

The system implements a feedback mechanism where the executed operation can be corrected by subsequent user utterances. The correction identification module detects when a user wants to correct the previous operation and processes the correction accordingly, allowing the system to learn from user feedback and improve accuracy over time

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The system performs preliminary action by executing the highest confidence command immediately, then prepares to handle potential corrections. The correction handling mechanism is pre-configured to efficiently process correction utterances, balancing speed with accuracy by having the correction pathway ready in advance

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If the system allows correction of operations, then the accuracy of user intent interpretation is improved, but the complexity of the system increases

Engineering Contradiction:
Improveuser intent interpretation accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the speech recognition process into distinct modules: command recognition module, operation execution module, correction identification module, and correction processing module. This segmentation allows each module to specialize in one function, making the overall complex system manageable and maintainable while improving accuracy

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The correction identification module uses the same command recognition capabilities as the initial command processing, making the system multi-functional. The same confidence score calculation and command matching logic serves both initial recognition and correction scenarios, reducing the need for separate complex mechanisms

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Reliability

If the system processes correction utterances, then the quality of speech recognition service is improved, but the response time increases

Engineering Contradiction:
Improveservice qualityVSAvoidresponse time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system dynamically adjusts its processing behavior based on whether a correction is detected. When a correction utterance is identified, the system switches to correction processing mode, which re-evaluates commands excluding the previously executed one. This dynamic adaptation allows the system to maintain high service quality while minimizing unnecessary processing time for non-correction scenarios

Inventive Principle:
Principle #15Dynamics

Data Source

PatentUS20240071383A1Method for analyzing user utterance and electronic device supporting the same
Publication Date: 2024.02.29 SAMSUNG ELECTRONICS CO LTD
  • US20240071383A1 patent drawing
  • US20240071383A1 patent drawing
  • US20240071383A1 patent drawing

AI summary

According to an embodiment, an electronic device comprises a microphone and at least one processor operatively connected with the microphone, wherein the processor is configured to obtain one or more commands corresponding to a first utterance obtained through the microphone, based on the first utterance, execute an operation corresponding to a first command having a highest confidence score for the first utterance among the one or more commands, identify that a second utterance obtained through the microphone is an utterance for correcting the operation corresponding to the first command, based on the second utterance, determine a second command having another highest confidence score for the second utterance among the one or more commands except for the first command, based on identifying that the second utterance is the utterance for correcting the operation corresponding to the first command, and execute another operation corresponding to the second command that is determined.