Offline Grammar Model for Voice Query Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Voice-based user interfaces on resource-constrained devices face challenges with continuous online connectivity requirements, limiting their functionality in offline modes, especially in mobile and automotive applications.
Innovation Solution
An offline semantic processor uses a dynamically generated and updated offline grammar model based on query usage data to parse voice-based queries, reducing resource requirements and enabling offline processing by incorporating popular queries from an online grammar model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a full online grammar model is deployed on resource-constrained devices, then query recognition accuracy is improved, but device memory and processing resources are overwhelmed
Solution Approach 1:
The grammar model is segmented into multiple components: a compressed offline grammar model for basic query recognition, and a separate online grammar model for complex queries. The offline model contains only essential grammar rules needed for offline operation, while the online model provides comprehensive coverage when connectivity is available. This segmentation allows the device to operate with limited memory resources while maintaining the option to access full functionality online.
Solution Approach 2:
The grammar model undergoes parameter changes through compression techniques that reduce its size for offline deployment. The offline grammar model uses simplified grammar rules and reduced vocabulary coverage compared to the full online model, allowing it to fit within device memory constraints while still providing adequate query recognition for common operations.
2Quantity of substance
If a compressed offline grammar model is used, then device resource requirements are reduced, but query recognition coverage is limited
Solution Approach 1:
An online service acts as an intermediary to supplement the limited offline grammar model. When the device is online, the offline grammar model handles initial query processing, and complex or unrecognized queries are forwarded to the online service for comprehensive processing. The online service can access the full grammar model and return results to the device, effectively extending the query recognition coverage beyond what the compressed offline model alone could provide.
Solution Approach 2:
The system dynamically adapts between offline and online modes based on connectivity and query complexity. The offline grammar model provides immediate response for common queries without requiring full model loading, while the system can transition to online processing for complex queries. This dynamic approach allows the system to maintain versatility across different operating conditions while respecting resource constraints.
3Power
If voice-based user interface processing is performed online, then processing capability is improved, but continuous internet connectivity is required
Solution Approach 1:
The essential grammar rules and vocabulary are pre-loaded into a compressed offline grammar model during device setup or through periodic updates when online. This preliminary action ensures that the device has the necessary processing capability stored locally to handle voice queries without requiring continuous internet connectivity. The offline model is prepared in advance to provide immediate processing capability when needed.
Solution Approach 2:
A simplified copy of the online grammar model is created for offline use. The offline grammar model contains a subset of the grammar rules and vocabulary from the full online model, optimized for size and speed. This copy provides sufficient processing capability for common queries and basic operations, ensuring reliable offline functionality while the full online model remains available for when connectivity is restored.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
An offline semantic processor of a resource-constrained voice-enabled device such as a mobile device utilizes an offline grammar model with reduced resource requirements to parse voice-based queries received by the device. The offline grammar model may be generated from a larger and more comprehensive grammar model used by an online voice-based query processor, and the generation of the offline grammar model may be based upon query usage data collected from one or more users to enable a subset of more popular voice-based queries from the online grammar model to be incorporated into the offline grammar model. In addition, such a device may collect query usage data and upload such data to an online service to enable an updated offline grammar model to be generated and downloaded back to the device and thereby enable a dynamic update of the offline grammar model to be performed.