Menu
▾
▴
#302 different results from hfst-proc and hfst-optimised-lookup
$ echo Eurohpálaš|hfst-proc analyser-mt-apertium-desc.nob.hfstol
^Eurohpálaš/Eurohpá<n><prop><der_laš><attr>/Eurohpá<n><prop><der_laš><sg><nom>/Eurohpá<n><prop><sem_plc><der_laš><attr>/Eurohpá<n><prop><sem_plc><der_laš><sg><nom>/Eurohpálaš<attr>/Eurohpálaš<sg><nom>$
$ echo Eurohpálaš|hfst-optimised-lookup analyser-mt-apertium-desc.nob.hfstol
Eurohpálaš +?
This is revision 116626 of
https://victorio.uit.no/langtech/trunk/langs/sme/
where the analyser is in tools/mt/apertium
configured with
./autogen.sh && ./configure --enable-apertium --with-hfst --without-xfst --disable-spellers --disable-analysers --disable-generators
fstol
$ hfst-strings2fst | hfst-compose -F proc-ol-differs.hfst | hfst-fst2txt
Eurohpálaš
0 1 @U.Cap.Obl@ @U.Cap.Obl@ 0.000000
1 2 E E 0.000000
2 3 u u 0.000000
3 4 r r 0.000000
4 5 o o 0.000000
5 6 h h 0.000000
6 7 p p 0.000000
7 8 á á 0.000000
8 9 @U.Cap.Opt@ @U.Cap.Opt@ 0.000000
9 10 l <n> 0.000000
10 11 a <prop> 0.000000
11 12 š @0@ 0.000000
11 12 š <sem_plc> 0.000000
12 13 @0@ <der_laš> 0.000000
13 14 @0@ 0.000000
14 15 @0@ <sg> 0.000000
14 16 @0@ <attr> 0.000000
15 17 @0@ <nom> 0.000000
16 18 @D.CmpOnly.FALSE@ @D.CmpOnly.FALSE@ 0.000000
17 18 @D.CmpOnly.FALSE@ @D.CmpOnly.FALSE@ 0.000000
18 19 @D.CmpPref.TRUE@ @D.CmpPref.TRUE@ 0.000000
19 20 @D.NeedNoun.ON@ @D.NeedNoun.ON@ 0.000000
20 0.000000
OK, this can be closed, it's the expected behaviour.
This fst was supposed to be composed with one that disallowed the der_laš tag, so I thought there must be a bug when the disallowing "worked" with hfst-optimised-lookup but not hfst-proc. But in fact, the composition never happened, and the only reason is that hfst-proc allows titlecased analyses but hfst-optimised-lookup doesn't, so
$ echo eurohpálaš|hfst-optimised-lookup analyser-mt-gt-desc.hfstol
eurohpálaš Eurohpá+N+Prop+Sem/Plc+Der/laš+A+Sg+Nom
eurohpálaš Eurohpá+N+Prop+Sem/Plc+Der/laš+A+Attr
eurohpálaš eurohpálaš+A+Sg+Nom
eurohpálaš eurohpálaš+A+Attr
sorry for the noise …