Voice Command Extraction From Conversational Utterances

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing devices struggle to accurately extract and execute commands from natural user utterances without requiring a wake-up word or specific command, limiting their functionality in conversational contexts.

Innovation Solution

A method and computing apparatus utilize a language model trained to identify target texts within user utterances, distinguishing between command-related and non-command-related text to control devices effectively.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a wake-up word or specific command is required before voice input, then voice recognition accuracy is improved, but device complexity and ease of operation deteriorate

Engineering Contradiction:

Improvevoice recognition accuracyVSAvoidease of operation

Core Design Contradiction:

Measurement precisionVSEase of operation

Solution Approach 1:

The system performs preliminary action by continuously monitoring for wake-up words to activate voice recognition functionality before actual commands are processed. This resolves the contradiction by preparing the system in advance without requiring complex manual intervention from users.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The wake-up word acts as an intermediary signal that bridges the gap between ambient noise and actual voice commands. It serves as a mediator to distinguish relevant speech from background noise, improving recognition accuracy while maintaining simple user interaction.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Ease of operation

If natural language understanding is used to extract commands from conversational utterances, then ease of operation is improved, but reliability deteriorates due to difficulty in distinguishing commands from conversation

Engineering Contradiction:

Improveease of operationVSAvoidcommand extraction reliability

Core Design Contradiction:

Ease of operationVSReliability

Solution Approach 1:

The system segments the utterance into potential command portions and non-command portions by analyzing linguistic patterns and context. This segmentation allows the system to distinguish between actual commands and conversational filler, improving both ease of operation and reliability simultaneously.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system uses feedback mechanisms to continuously refine its understanding of user intent by analyzing the context and structure of utterances. This feedback loop enables the system to learn from previous interactions and improve command extraction accuracy over time.

Inventive Principle:

Principle #23Feedback

3Reliability

If rule-based NLU is used to distinguish commands from conversation, then reliability is improved, but adaptability deteriorates

Engineering Contradiction:

Improvecommand extraction reliabilityVSAvoidadaptability

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The system changes parameters dynamically based on context, adjusting its command detection thresholds and criteria according to the conversational situation. This allows the system to maintain high reliability through adaptive parameter adjustment rather than fixed rules, thereby improving both reliability and adaptability.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20260031086A1Method for controlling device on basis of command extracted from user utterance and computing apparatus for performing same

Publication Date: 2026.01.29 SAMSUNG ELECTRONICS CO LTD

AI summary

A method is provided. The method includes obtaining an utterance of a user, determining whether a target text requesting a device to perform a function is included in the utterance by using a language model, and controlling the device based on the target text based on determining that the target text is included, wherein the language model is a model trained to extract text related to a request from successive sentences.