Voice Intent Classification via Multi-Attribute Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional electronic devices face limitations in determining whether a user's current utterance is a subsequent utterance candidate, as they rely solely on time interval information, which is insufficient to accurately assess the intent and domain consistency of user voices.
Innovation Solution
An electronic device and control method that classify multiple attributes of a user's voice, including time, utterance frequency, device state, speaker identity, and command similarity, and apply different weights to these attributes to determine if the voice is a subsequent utterance candidate, thereby improving the accuracy of intent and domain determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If electronic devices use only time interval information to determine subsequent utterances, then the determination process is simple and fast, but the accuracy of intent and domain determination is insufficient
Solution Approach 1:
The patent segments the determination process into multiple independent attribute classifications (time attribute, speaker attribute, command attribute, device state attribute, utterance frequency attribute). Each attribute is evaluated separately and then combined to make the final determination, allowing the system to achieve high accuracy through multiple dimensions while maintaining modular complexity
Solution Approach 2:
The patent transitions from a single-dimensional time-based determination to a multi-dimensional attribute-based determination system. By adding dimensions such as speaker identity, command similarity, device state, and utterance frequency, the system achieves more accurate intent determination without overcomplicating the overall structure
2Measurement precision
If electronic devices consider multiple attributes for determining subsequent utterances, then the accuracy of intent determination is improved, but the resource burden on voice recognition systems increases
Solution Approach 1:
The patent applies partial action by selectively evaluating attributes based on their relevance to the current context. Not all attributes are evaluated with equal intensity - the system adjusts the depth of analysis for each attribute type, applying more detailed analysis to critical attributes like speaker identity and command similarity while using lighter evaluation for less critical attributes
Solution Approach 2:
The patent changes the parameters of attribute evaluation dynamically. Different weights are assigned to different attributes based on their importance, and the evaluation threshold for each attribute can be adjusted. This allows the system to optimize resource usage by focusing computational effort on attributes that provide the most value for accurate intent determination
Data Source
AI summary
The present disclosure provides an electronic device and a control method therefor. The electronic device of the present disclosure comprises: a voice reception unit; and a processor for, when a first user voice and a second user voice are received through the voice reception unit, determining whether the second user voice corresponds to a candidate of utterance subsequent to the first user voice on the basis of a result obtained by dividing a plurality of attributes of the second user voice according to a predefined attribute, and controlling the electronic device to perform an operation corresponding to the second user voice on the basis of the intent of the second user voice obtained through a result of the determination.