Parallel Speech Recognition Engines with Context Database
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition systems using a single search-recognition engine are inflexible and reduce accuracy, often failing to recognize spoken utterances, leading to error conditions and reduced usefulness in directory-assistance applications.
Innovation Solution
A system comprising a directed-dialog-processor server, a speech-recognition-engine server with multiple parallel-operable speech-recognition-engine applications, and a multiple-recognition-processor server that selects and combines context information to perform speech recognition, generating an n-best list with confidence-score and word-score values to improve recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single search-recognition engine is used for speech recognition, then the system complexity is low, but the recognition accuracy and flexibility are reduced
Solution Approach 1:
The system divides the speech recognition task into multiple parallel search-recognition engines, each specializing in different recognition approaches (context-dependent, context-independent, out-of-vocabulary). This segmentation allows each engine to optimize for its specific function while collectively providing comprehensive recognition coverage, resolving the contradiction between accuracy and complexity.
Solution Approach 2:
The system merges the results from multiple parallel search-recognition engines by combining their confidence scores and selecting the highest-scoring recognition result. This combining approach integrates the strengths of different recognition engines to achieve higher overall accuracy while managing system complexity through coordinated result aggregation.
2Adaptability or versatility
If multiple parallel speech-recognition engines are used, then recognition accuracy and flexibility improve, but system complexity increases
Solution Approach 1:
The system implements multiple search-recognition engines that can handle different types of speech inputs (context-dependent, context-independent, out-of-vocabulary) within a unified framework. Each engine provides specialized functionality while the overall system maintains universal applicability across various speech recognition scenarios, enhancing flexibility without proportionally increasing complexity.
Solution Approach 2:
The system uses confidence scores from each search-recognition engine as feedback to determine which engine's result to accept. The highest confidence score determines the final recognition outcome, and this feedback mechanism allows the system to dynamically select the most appropriate recognition result, improving flexibility while managing complexity through automated decision-making.
3Reliability
If a single search-recognition engine is used, then the system is simpler to operate, but error conditions increase and usefulness decreases
Solution Approach 1:
The system prepares multiple search-recognition engines in advance, each specialized to handle different types of speech recognition challenges. By having these engines ready beforehand with different expertise (context-dependent, context-independent, out-of-vocabulary), the system cushions against recognition errors that would occur with a single engine, improving reliability while maintaining ease of operation through automated error prevention.
Data Source
AI summary
A method of and system for accurately determining a caller response by processing speech-recognition results and returning that result to a directed-dialog application for further interaction with the caller. Multiple speech-recognition engines are provided that process the caller response in parallel. Returned speech-recognition results comprising confidence-score values and word-score values from each of the speech-recognition engines may be modified based on context information provided by the directed-dialog application and grammars associated with each speech-recognition engine. A context database is used to further reduce or add weight to confidence-score values and word-score values, remove phrases and/or words, and add phrases and/or words to the speech-recognition engine results. In situations where a predefined threshold-confidence-score value is not exceeded, a new dynamic grammar may be created. A set of n-best hypotheses of what the caller uttered is returned to the directed-dialog application.