On 8 August 2018 at 10:36, Xavi Ivars <xavi.iv...@gmail.com> wrote:
> If that is not the root cause, we need to find out what the root cause is.
> But having two versions of the tagger that, with exactly the same input and
> the same prob files does not generate the same output is really bad.
>
> I double checked, and th ecoarse tags for the tagger do distinguish
> between both input words (np.loc and np.ant). Is there anyone that can help
> a bit on debugging apertium-tagger? Things like displaying the
> probabilities for each option would help us identify if this is just
> something that is a coincidence (both tags have exactly the same probs) or
> if there's something wrong.
>
apertium-tagger has always had random segfaults, well before weight code
was added. I wouldn't be surprised if there's UB leading to randomness
somewhere in there.
The lttoolbox segfault is now fixed and passes its own "make test". I'll
investigate tagger further.
-- Tino Didriksen
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Apertium-stuff mailing list
Apertium-stuff@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/apertium-stuff
