[libvoikko] Sámi/HFST

[Harri Pitkänen]

On Monday 28 June 2010, Harri Pitkänen wrote:
> > If you are still missing the error models, I tossed the transducer
> > version of error model in hunspell se_FI.aff file of ubuntu
> > distribution to <http://www.helsinki.fi/~tapirine/tmp/se_FI.err.hfst>.
> 
> OK, thanks. I was away for two weeks but will return to this soon and see
> if I  can get similar suggestions from this and Hunspell

I was able to build a transducer that produces suggestions similar to 
Hunspell. For example with randomly picked input string "ramus":

$ echo ramus | hfst-lookup sug.hwfst
ramus   gamus
ramus   hamus
ramus   lamus
ramus   oamus
ramus   ragus
ramus   ramas
ramus   raŋus
ramus   rašus
ramus   romus
ramus   samus

$ echo ramus | hunspell -d se 2>/dev/null
Hunspell 1.2.11
& ramus 14 0: ramsu, áramus, rampus, ramsus, rašus, romus, oamus, raŋus, 
lamus, hamus, ragus, gamus, samus, ramas

But this does not work with libvoikko. No suggestions are produced even if the 
backend is configured just the same way as with Omorfi. The code in 
HfstSuggestion.cpp seems to assume something about the internals of the 
suggestion transducer but I can't easily figure out what the difference is. I 
did use hfst-summarize with both transducers and could not see any significant 
structural difference between them.

Harri