http://jakarta.apache.org/poi/index.html

To do better, maybe you can get word format specs under NDA from
Microsoft.  Good luck with that.  I'm not being *too* cynical here.
I've worked for a company and for a university that each had a certain
amount of Microsoft source code, so who knows?

Maybe there's a Word Doc to PDF converter, at least that would give you
a known format.  

Maybe there's a way to connect JNI to Winelib and use some Office DLL to
do the conversion.  Or, connect to one of the commercial printer drivers
that converts Word to Postscript or PDF, and deal with the output.  Or,
maybe you can connect an ActiveX control that can display a Word
document, and then screen scrape [I'm serious.]

But you are treading into territory where Microsoft has gone to great
lengths to prevent you from doing what you need to do.  If your
documents are in Word97 doc format, and don't have too much (any) OLE or
weird formatting.

The easiest way might be to set up an old clunker under Windows, and
write a little program on it that wakes up every once in a while
looking for files in its queue, and it is use EXEC to feed them to
word itself for printing. Your server then only has to put the files
in a directory on disk. It can once a day clear files more a day old,
or allow the windows program to delete once successfully printed.