This could be implemented in a similar way to the Location Services control, in that when emplaced on a view, it could support an action 'Start Listening' and that point, the data could be passed to WKWebView so that it can then be handed to an external service for NLP or use the browser's Web Speech API to seek specific words.

https://stackoverflow.com/questions/43916436/enable-camera-and-mic-access-in-wkwebview