I was playing around with TLSphinx (Swift wrapper for Pocketsphinx) and gave up after a while. It seemed ideal for my limited grammar use case. Instead I'd like to use the off-grid version of SFSpeechRecognizer, but I'm concerned about recognizing domain specific terminology and the ambiguity of the enormous large language model. Is there a way to customize SFSpeechRecognizer in such a way to limit its grammar?
Asked
Active
Viewed 52 times
