Anti NLP Brief

Natural language processing

Academic way, which is knowledge-intensive and very scientific, is too heavy. Parsing sentences just for extracting parts of speech and what?
Getting "persons" like "he" and "she" from sentence with coreNLP does not seem very useful.
It needs a lot of dictionaries, a lot of memory and CPU time for processing. i don't see how it is applicable for real tasks, for example:

I would like to compare mapping of events in different times, described in different cultures. Might it be there some intersection? I mean similar events sequence described in different places and times?

I think, for this Epic I could try individual approach of application level. Of course, I need to prepare data - parse several ancient sources - Sumerian, Egyptian, Greek, Roman historic texts. I need to get dates in any form (ages, centuries, years, days) and build timeline with context.
I need to pick persons and places, then try to slide one set against another and, who knows, may be some intersection will be found?