Adapting Local Speech Recognition Grammar via Server Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Local speech recognition engines typically have inferior performance compared to server-based systems and are less frequently updated, while server-based systems are not always available due to connectivity issues, especially in remote areas.
Innovation Solution
A framework that adapts local speech recognition systems by leveraging server-based systems, using a processor to communicate between local and server-based engines, storing recognition results, and incorporating high-confidence user-confirmed inputs into the local engine's grammar for improved performance, allowing local operation when server connectivity is lost.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If local speech recognition engine is used, then availability is guaranteed, but performance is inferior
Solution Approach 1:
The system performs preliminary actions by collecting speech inputs and their recognition results from the server-based engine, storing them in a database. When confidence scores exceed thresholds and users confirm correctness, these results are used to adaptively update the local engine's grammar and models in advance, so that when operating independently, the local engine already has improved recognition capabilities.
Solution Approach 2:
The system implements feedback mechanisms where recognition results from the server-based engine are used to update and adapt the local speech recognition engine. The local engine continuously learns from the server's high-quality recognition results, creating a feedback loop that improves local performance over time while maintaining availability.
2Measurement precision
If server-based speech recognition engine is used, then performance is superior, but availability is reduced due to connectivity issues
Solution Approach 1:
The system introduces an intermediary adaptive layer that translates and adapts the server-based engine's recognition results into formats suitable for the local engine. This intermediary process allows the local engine to benefit from server-quality performance through adapted grammar and models, enabling it to operate independently when connectivity is unavailable.
Solution Approach 2:
The system performs preliminary adaptation of server recognition results into local engine-compatible formats, storing adapted grammar and models in advance. This preliminary action ensures that when the server becomes unavailable, the local engine already possesses adapted capabilities to maintain superior performance without real-time connectivity.
3Reliability
If local speech recognition engine is used, then availability is maintained, but update frequency is lower
Solution Approach 1:
The system establishes a feedback mechanism where the local engine continuously receives update information from the server-based engine. When server recognition results meet confidence thresholds and receive user confirmation, these updates are automatically transmitted to adapt the local engine, creating a continuous improvement cycle that increases update frequency while maintaining availability.
Solution Approach 2:
The system performs preliminary collection and validation of recognition results from the server, preparing update packages in advance. This preliminary action enables more frequent and efficient updates to the local engine, as the adaptation work is partially completed before transmission, increasing update frequency without compromising availability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A speech recognition method includes providing a processor communicatively coupled to each of a local speech recognition engine and a server-based speech recognition engine. A first speech input is inputted into the server-based speech recognition engine. A first recognition result from the server-based speech recognition engine is received at the processor. The first recognition result is based on the first speech input. The first recognition result is stored in a memory device in association with the first speech input. A second speech input is inputted into the local speech recognition engine. The first recognition result is retrieved from the memory device. A second recognition result is produced by the local speech recognition engine. The second recognition result is based on the second speech input and is dependent upon the retrieved first recognition result.