Solved the API compatibility issue with the Recognizer interface. Now both Web Speech and Azure implement the interface and expose the same few public methods to be used by useRecognition.
Further, refactored transcript stream to a list of finalized transcript blocks and a single in-progress transcript block to allow all APIs to yield the same transcript updates. Unfortunately this meant changing Tiger's data structure quite drastically - please review this part if you could!
Solved the API compatibility issue with the Recognizer interface. Now both Web Speech and Azure implement the interface and expose the same few public methods to be used by useRecognition.
Further, refactored transcript stream to a list of finalized transcript blocks and a single in-progress transcript block to allow all APIs to yield the same transcript updates. Unfortunately this meant changing Tiger's data structure quite drastically - please review this part if you could!