[libvoikko] Setting up HFST morphology backend

[Flammie Pirinen]

2010-04-27, Sjur Moshagen sanoi:

> Den 26. apr. 2010 kl. 18.18 skrev Flammie Pirinen:

> > For HFST backend you can cp ${prefix}/share/omorfi/*
> > ${HOME}/.voikko/2/mor-hfst/ and rename stuff like so:
> > 
> > ls ~/.voikko/2/mor-hfst/
> > hyphenation.hwfst  hyp.hwfst  mor.hwfst  spl.hwfst  sug.hwfst
> > voikko-fi_FI.pro
> 
> Does that mean that all transducers used by voikko should be
> weighted? Presently, the make target in omorfi only produces
> unweighted transducers, at least based on the filename suffixes.

Yes, currently voikko's backend uses weighted transducers IIRC, the
omorfi also makes weighted ones by default, but the distinction has
been dropped from file extensions in wait of forthcoming HFST versions,
since the transducers will now more or less work interchangingly and it
would've not made sense to encode possible backends used in the
filename, since amount of backends seems to be growing as well.

Of course in a few days or weeks the HFST backend of voikko should
switch to optimized format instead, since I already have most parts of
the code done and it's much nicer than the current hacks.

-- 
Flammie, computer scientist bachelor, linguist master, free software
Finnish localiser, and more! <http://www.iki.fi/flammie/>