Voice Recognition Model Management via Usage-Based Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing natural language understanding models for electronic devices require significant storage capacity and can take time to download, especially when not connected to a network, as they need to be stored for all applications and updated separately for each application, leading to inefficiencies in voice recognition and user interaction.
Innovation Solution
An electronic apparatus that acquires and manages a natural language understanding model based on usage information, such as execution frequency and time, to efficiently store and update models, allowing for pre-emptive downloading of models for frequently used applications, using both personal and public models for voice recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If natural language understanding models are stored for all applications, then voice recognition accuracy is improved, but storage capacity requirements increase
Solution Approach 1:
The patent segments the natural language understanding models by application, storing them separately in a database rather than requiring all models to be simultaneously available in memory. The system divides models into multiple categories (public models, personal models, application-specific models) and retrieves only the necessary segments based on usage patterns.
Solution Approach 2:
The system performs preliminary actions by pre-acquiring and storing natural language understanding models in a database before they are needed for voice recognition. Usage information is collected in advance to determine which models to acquire, and models are stored ready for retrieval when voice input occurs, eliminating the need for real-time downloading during voice processing.
2Adaptability or versatility
If natural language understanding models are downloaded from server, then model availability is improved, but time delay increases and network dependency increases
Solution Approach 1:
The system performs preliminary acquisition of natural language understanding models and stores them locally in a database before they are needed. Usage information is analyzed in advance to determine which models to acquire, and these models are stored ready for immediate retrieval when voice input occurs, eliminating the need for real-time downloading during voice processing.
Solution Approach 2:
The patent introduces a local database as an intermediary between the server and the voice recognition processing. Instead of directly downloading models from the server during voice processing, the system uses the local database to store and retrieve models, acting as a buffer that eliminates network dependency during actual voice recognition operations.
3Measurement precision
If natural language understanding models are updated for each application, then recognition accuracy is improved, but system complexity increases
Solution Approach 1:
The patent creates a universal natural language understanding model database that serves multiple applications simultaneously. Instead of maintaining separate model management systems for each application, the system uses a single database structure that can store and retrieve models for different applications based on usage information, reducing overall system complexity while maintaining application-specific accuracy.
Solution Approach 2:
The system implements self-service by automatically analyzing usage information to determine which natural language understanding models need to be acquired and stored. The process autonomously identifies frequently used applications, retrieves appropriate models, and updates the database without requiring manual intervention, simplifying the model management process while maintaining high recognition accuracy.
4Speed
If all natural language understanding models are stored locally, then voice recognition speed is improved, but storage requirements increase
Solution Approach 1:
The patent applies local quality by storing natural language understanding models locally in a database based on their specific usage patterns and importance. Frequently used application models are stored locally for immediate retrieval and fast voice recognition, while less frequently used models can be managed differently. This selective local storage optimizes both speed and storage efficiency.
Solution Approach 2:
The system performs preliminary acquisition and local storage of natural language understanding models based on usage information analysis. By identifying which models will be needed and storing them in advance in a local database, the system ensures fast retrieval during voice recognition without requiring all possible models to be permanently stored, thus optimizing both speed and storage requirements.
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
An electronic apparatus is provided. The electronic apparatus includes: a memory configured to store at least one instruction; and a processor configured to execute the at least one instruction to: obtain usage information on an application installed in the electronic apparatus, obtain a natural language understanding model, among a plurality of natural language understanding models, corresponding to the application based on the usage information, perform natural language understanding of a user voice input related to the application based on the natural language understanding model corresponding to the application, and perform an operation of the application based on the preformed natural language understanding.