to your second Question: This is one JSON Object with three values:
{
"tagFirstValue" : "first value here",
"tagSecondValue" : "second value",
"tagthirdVale" : "third value"
}
Instead you could also use an array:
{"tag": [ "firstVale", "secondValue", "thirdValue"]}
Or use nested Objects:
{
"tagFirstObject" : {"firstObjectTag":"first value here"},
"tagSecondObjec" : {"secondObjectTag":"second value here"},
"tagthirdObjec" : {"thirdObjectTag":"third value here"},
}
Its up to you how you define your schema.
In the same way you should create a collection that fit to your RDF Schema, so this is to Question one.
to Question three: I think the best will be to use a mongodb-driver for the programming language of your choise and transform the SPARQL Query into a MongoDB query.