FIELD: physics, computer engineering.
SUBSTANCE: invention relates to speech recognition. Speech recognition employs an overall grammar as a set of context-specific grammars for recognition of an input data, each responsible for a specific context. The grammars together cover the entire domain. Multiple recognitions can be used in parallel with respect to the same input, where each recognition uses context-specific grammars. Multiple intermediate recognition results from different recogniser-grammar pairs are reconciled by running re-recognition using a dynamically composed grammar based on the multiple recognition results and other domain knowledge, or selecting the winner using a statistical classifier operating on classification features extracted from the multiple recognition results and other domain knowledge.
EFFECT: shorter delay during speech recognition, higher accuracy of speech recognition, larger overall grammatical coverage in speech recognition.
22 cl, 9 dwg
