1. All of the space characters in the hypothesis have been replaced with ‘#^#’
OK, I will check into that tomorrow. It seems like a regression.
2. It seems as if rapid ears guesses a complete statement halfway through hearing it. Is this the correct behavior? Is there some way to return only the part of the statement rapid ears has heard? Here is an extreme example I found while speaking “Greetings robot instruction go 20 left thank you”:
This is expected – a grammar means that nothing is possible outside the ruleset, so there can’t be any hypotheses that are just parts of rulesets. But on the UI/app design side of things it’s more expected with RapidEars that you’ll want to recognize much shorter phrases, so that there is a benefit to using realtime recognition, rather than trying to recognize really long ruleset phrases like the ones in the stock OpenEars grammar example.