Probably a "too simple" algorithm.
Their goal seems to be to build a concordance of all books on the machine.
That can be done (and maintained) "one book at a time".
Obviously, that is not the algorithm Amazon is using (your guess may be as close as any other).
I am sure we have the skills at this web-site to fix it for them, once they send us their sources.

But since "code for free" has no value to a corporation - we need to quote them a (big) price for the work.