Database Speech Recognition Resolution via Phonetic Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Standard Automated Speech Recognition (ASR) engines have limited resolution capabilities, particularly with large grammars, leading to inefficiencies in speech recognition applications, and existing solutions like specialized hardware and software systems face issues with data centralization, portability, and data synchronization.
Innovation Solution
Integrating large-scale speech recognition technology directly into the database layer, allowing voice access to data through a 'soundslike' function that converts audio requests into searchable phonetic forms, enabling efficient matching and returning ordered results without the need for specialized voice recognition engines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If standard ASR engines are used for speech recognition, then the system is simple and easy to implement, but the resolution capability is limited and cannot handle large grammars effectively
Solution Approach 1:
The patent introduces a database as an intermediary component between the ASR engine and the application. The database stores pre-compiled phonetic representations of grammar elements and serves as a centralized repository that both the ASR engine and application can access. This mediator enables the ASR engine to resolve large grammars by looking up phonetic matches in the database without requiring the engine itself to handle the complexity of large-scale grammar management.
Solution Approach 2:
The system divides the speech recognition functionality into separate components: the ASR engine handles audio processing and phonetic matching, while the database stores and manages the grammar data. This segmentation allows each component to be optimized independently - the ASR engine for recognition accuracy and the database for efficient data retrieval and storage of large grammar sets.
2Measurement precision
If specialized large-scale speech recognition technology is used, then the resolution capability for large grammars is improved, but data synchronization problems arise when data is replicated
Solution Approach 1:
The patent merges the grammar storage functionality into the existing database infrastructure rather than using a separate specialized system. By storing phonetic representations of grammar elements in the same database that the application uses, the system eliminates data replication and synchronization issues. The database serves dual purposes: storing application data and storing speech recognition grammar data, ensuring data consistency across the system.
3Measurement precision
If specialized speech recognition systems are deployed, then recognition accuracy for large vocabularies is improved, but the system becomes non-portable and locked into proprietary implementations
Solution Approach 1:
The patent creates a universal speech recognition system that works with standard database interfaces and can be implemented across different platforms. By using a database with standard query interfaces to store and retrieve phonetic data, the system can be deployed with different ASR engines and database systems without requiring proprietary implementations. The database layer provides a platform-independent interface that enhances system portability and adaptability.
4Quantity of substance
If grammar data is stored in a centralized location, then data centralization is improved and duplication is reduced, but data access speed may be reduced due to retrieval overhead
Solution Approach 1:
The system performs preliminary processing by pre-compiling and storing phonetic representations of grammar elements in the database before they are needed for speech recognition. This advance preparation allows the ASR engine to quickly retrieve and match phonetic data during runtime without performing complex processing operations, thereby maintaining fast access speeds while benefiting from centralized data storage.
Data Source
AI summary
A method of providing speech recognition resolution in a database includes receiving an utterance from an end-user or application, and dispatching it to the database which is coupled to or has access to a speech recognition technology. The method further includes converting the utterance to an intermediate form suitable for searching, and transmitting the intermediate form from the database to the speech recognition technology. An item-matching is then performed via the speech recognition technology which returns items matching the utterance to the database along with a confidence metric, and the utterance matches are then provided from the database in the form of an ordered result set.