Apart from the case in which a ban phrase is a subset of another longer ban phrase, there should also be an explanation about the case in which a ban phrase isn't a subset of another ban phrase, yet share common words.

For example, what should be the answer for below situation?

banPhrases = ["as long", "long as"]
article = "as long as"

Maybe consider reducing the random test input size a bit; it's big enough to cause problems printing the input of all test cases, and they certainly don't need to be that big. It's also too big to debug reasonably.