I don't see how NLP would be used in this fashion. NLP attempts to detect sentiment, language use, etc. Its results are presented with a "measured" confidence ... E.g. 90% confidence in analysis. There are "text similarity" uses but again - the end result is not a character by character comparison.
I see nothing in Pega's implementation of NLP that would fit this use case.