Call Center Voice Recognition Engine Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing call center voice recognition systems fail to select the most appropriate voice recognition engine based on language and specialized dictionary, leading to suboptimal recognition rates.
Innovation Solution
A call center system that includes a voice recognition server with a business information-recognition engine correspondence table, allowing for the selection of the appropriate voice recognition engine for each business, language, and type of business, thereby enhancing recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single voice recognition engine is used for all call types, then the system structure is simple, but the recognition rate decreases for specialized business calls
Solution Approach 1:
The patent applies parameter changes by selecting different voice recognition engines based on the type of call. The system changes the engine parameter (which recognition engine to use) according to the business category of the call, thereby optimizing recognition accuracy for different specialized domains without requiring a completely different system architecture.
Solution Approach 2:
The system implements dynamics by making the voice recognition engine selection dynamic rather than static. The appropriate engine is selected at runtime based on the call type identified from the incoming call data, allowing the system to adapt to different recognition requirements for different business scenarios.
2Measurement precision
If multiple voice recognition engines are maintained for different businesses, then the recognition rate increases, but the system complexity and management difficulty increase
Solution Approach 1:
The system applies self-service by automatically selecting the appropriate voice recognition engine based on the call type. The automatic selection mechanism eliminates the need for manual intervention to choose the correct engine, and the system self-manages the mapping between call types and engines through pre-configured correspondence tables.
Solution Approach 2:
The patent introduces an intermediary mechanism in the form of a correspondence table that maps call types to specific voice recognition engines. This intermediary structure simplifies the relationship between multiple engines and different call types, making the system easier to manage by providing a clear, organized mapping rather than direct complex connections.
3Measurement precision
If voice recognition is performed on all calls uniformly, then the processing flow is simple, but the recognition accuracy decreases for specialized terminology
Solution Approach 1:
The patent applies local quality by tailoring the voice recognition engine selection to the specific local requirements of each call type. Different business domains (e.g., medical, legal, technical support) have their own specialized terminology and patterns, and the system provides locally optimized recognition by selecting engines specifically suited for each domain rather than applying a uniform approach.
Data Source
AI summary
A call and recorded information management server transmits a requested call identification ID for recognition and an incoming call number corresponding to the ID, to a voice recognition server, which searches for the recognition engine corresponding to the number of the received call, and adds the received ID to a recognition queue. The voice recognition server requests the call and recorded information management server to obtain recorded data by the received ID. The call and recorded information management server transfers the recorded data corresponding to the ID to the voice recognition server to perform voice recognition by the corresponding recognition engine on the recorded data corresponding to the call identification ID stored in the recognition queue, and stores the result as text data. Thus, the most appropriate voice recognition engine can be used and the recognition rate of voice recognition can be increased.