Voice Interaction Cache for Local Semantic Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current voice interaction systems face challenges in balancing accuracy, efficiency, and cost, particularly when a limited number of servers provide voice interaction services to a large number of terminal devices, as they either require high processing and storage resources on devices or struggle with ensuring response speed and accuracy.
Innovation Solution
A terminal-device-based method that caches sample sentences and response content, performing local semantic matching and sending inputs to a server for understanding only when matches are not found, reducing server load and improving response speed without excessive resource demands.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all voice inputs are sent to the server for semantic understanding, then accuracy is maintained, but server processing pressure increases and response speed decreases
Solution Approach 1:
The system segments voice inputs into two categories: common sentences (matched against cached sample sentences) and uncommon sentences (sent to server for semantic understanding). This segmentation allows the majority of routine queries to be handled locally, reducing server load while maintaining accuracy for complex queries.
Solution Approach 2:
The system performs preliminary semantic matching against cached sample sentences before sending inputs to the server. This preliminary action filters out common queries that can be answered directly from cache, preventing unnecessary server processing and improving overall system efficiency.
2Speed
If semantic matching is performed locally on terminal devices, then response speed improves, but device processing and storage resources increase
Solution Approach 1:
The system creates a simplified copy of frequently used sample sentences and stores them locally on terminal devices. This copying approach enables fast local matching for common queries without requiring full semantic understanding capabilities on the device, balancing response speed with resource constraints.
Solution Approach 2:
The system implements local semantic matching capability on terminal devices specifically for common sample sentences, while relying on server-side processing for complex queries. This local quality approach provides fast response for routine operations without overburdening device resources with full AI processing capabilities.
3Measurement precision
If cached sample sentences are updated frequently, then interaction accuracy improves, but cache management complexity increases
Solution Approach 1:
The system implements feedback mechanisms where unmatched voice inputs are sent to the server, and successful server responses are used to update the local cache. This feedback loop continuously improves cache accuracy by incorporating real-world interaction data without requiring complex manual cache management.
Solution Approach 2:
The cache system automatically updates itself by receiving feedback from server interactions. When the server successfully processes an unmatched query, that sample sentence and its response are added to the local cache, enabling the system to self-improve without external intervention or complex management protocols.
Data Source
AI summary
The present disclosure discloses a voice interaction method, system, terminal device and medium. The method comprises the following operations performed by a terminal device: performing voice recognition on collected voice signals to acquire an input sentence; semantically matching the input sentence with cached sample sentences to determine whether there is a sample sentence having the same or similar semantics as the input sentence among the cached sample sentences; if yes, acquiring cached response content corresponding to the sample sentence having the same or similar semantics as the input sentence as response content of the input sentence; if not, sending the input sentence to a server; receiving from the server response content acquired by the server through semantic understanding according to a knowledge base stored on the server; and responding to the input sentence according to the response content.