Bug#403619: languagetool -- rule-based language checker
On Sun, May 13, 2012 at 04:35:36PM +0200, Daniel Naber wrote:
> Lib exists but is not up-to-date (I checked 'unstable'):
>
> libsegment-java 1.3.5, LT needs 1.3.0 and LT 1.8 will need 1.3.8
> libjwordsplitter-java 3.0, LT needs 3.3
> libmorfologik-stemming-java 1.2.2, LT needs morfologik-fsa-1.5.2 and
> morfologik-stemming-1.5.2 (the lib has been split up)
>
> Libs that I did not find in Debian and that we require:
>
> tika-core-0.9.jar from http://tika.apache.org/, Apache License 2.0
Don't forget dependencies of dependencies :)
> Libs that I did not find in Debian but that are only required for Chinese so
> I think we could do without for now:
>
> ictclas4j-1.0.jar from http://code.google.com/p/ictclas4j/,
> Apache License 2.0
> CJFtransform_v1.0.1_bin.jar from http://code.google.com/p/cjftransform/,
> Apache License 2.0
I'd "cripple" it only as last resort but first trying to package those.
> The internal dictionaries we use are huge when saved as text files (e.g.
> 200MB for German alone). Thus we compress them as a finite-state automaton
> with the morfologik-stemming project, which yields a 10 time better
> compression than bzip2 (tested with the German dictionary). We
> describe how to dump the dictionaries to plain text at the URL that Marcin
> has posted.
And how do you - which is the point after you did changes - create them (again)?
Is that also documented? Best would be some make-like thingy.
> The question is, what can we do now to help the process of getting LT into
> Debian?
Get all the packages updated and make the dictionaries be built during the package
build (see above).
And then someone needs to do the package and maintain it :-)
(And note not all architectures in Debian do have a Java 6[1] and you
sttill can choose gcj is you wish on the others, so you need to prevent
LT from trying to do something if LO is configured to use gcj)
Regards,
Rene
[1] kfreebsd-* only has gcj. No openjdk there.
Reply to: