Consider splitting out the Java and non-Java bits of the package

[richierocks]

rJava is a bit painful in that it sometimes takes a ridiculous amount of effort to get working on some systems.

I see that in previous issues there has been talk of moving away from rJava, but the xlsx and OpenNLP dependencies caused problems.

It seems like quite a lot of the package doesn't depend upon Java, so it would be useful to split this functionality out into a separate package without the dependency. See assertive and devtools for examples of how packages have been decomposed into underlying pieces.