Is it possible to give a text sentence and a sound sample to Mozilla-DeepSpeech software and get a matching percentage of both ?
It to make a quizz software. Program know the response and ask user to pronounce it in the microphone.
No, DeepSpeech will simply generate the best it can "hear" from the audio you provide. You could use it's test functionality to do that, but that is a bit of an overkill.
Try a tool like BenchmarkSTT for that.