Comment 3366

Date: 2010-06-17 01:39:01 +0200
From: Francis Tyers <>

Could we get one of the variants marked with a tag so that we can just generate one variant per analysis ?

$ echo "NSR oaivvilda ahte stuorra oassi guovttegielalašvuođadoarjagiin, mat leat juhkkojuvvon suohkaniidda ja fylkkasuohkaniidda, geavahuvvojit dakkár áigumušaide mat eai leat definerejuvvon njuolggadusain." | apertium -d . sme-smj
*NSR árvvet ahte #stuores oasse *guovttegielalašvuođadoarjagiin, ma/mahá li juohkám suohkanijda ja fylka suohkanijda, adni #dakkár @muš ma/mahá e la/le *definerejuvvon njuolgadusájs.

================================================================================

$ echo ma | osmj
0%>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>100%
ma mij+Pron+Interr+Pl+Nom
ma mij+Pron+Rel+Pl+Nom

$ echo mahá | osmj
0%>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>100%
mahá mij+Pron+Rel+Pl+Nom
mahá mij+Pron+Interr+Pl+Nom

================================================================================

mij+Pron+Interr:ma%> oblintercas ;
mij+Pron+Interr+Pl+Nom:ma%>há K ;

...

LEXICON K-s ! Alternative clitic lexicon
+Clt:s # ;

;

================================================================================

from lang-smj.

Comment 3367

Date: 2010-06-17 11:44:48 +0200
From: Trond Trosterud <<trond.trosterud>>

Šhat about having mahá as baseform for the latter?

from lang-smj.

Comment 3369

Date: 2010-06-17 12:25:37 +0200
From: Thomas Omma <<thomas.omma>>

use/NG ?

from lang-smj.

Comment 4388

Date: 2011-06-04 13:51:02 +0200
From: Trond Trosterud <<trond.trosterud>>

Yes, but it is not enough. I compiled smj with ifst-restr (the transducer I got was marked GG (the default dialect :-(, a bit misplaced since GG is a sme dialect, but this is only a cosmetic error), the resulting ismj-GG.restr.fst cut down the +Abe forms by 3/4, but still left 2 +Pl+Nom. So yes, +Use/NG works, and is the answer to the bug, provided that

• Apertium uses a generator fst which does not contain +Use/NG strings
• the +Use/NG string leaves us with one form, and not 2-4, as today.

It might be that Apertium already used the restrictive fst (I think so), the bug is in any case still open, also on our side.

from lang-smj.

Comment 4391

Date: 2011-06-04 14:00:13 +0200
From: Trond Trosterud <<trond.trosterud>>

Output with the restr fst:

~/main/gt$echo mij+Pron+Rel+Pl+Abe | lookup -flags mbTT
smj/bin/ismj-GG.restr.fst
0%>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>100%
mij+Pron+Rel+Pl+Abe majda
mij+Pron+Rel+Pl+Abe maj#da
mij+Pron+Rel+Pl+Abe maj#da
mij+Pron+Rel+Pl+Abe majda

~/main/gt$echo mij+Pron+Rel+Pl+Nom | lookup -flags mbTT
smj/bin/ismj-GG.restr.fst
0%>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>100%
mij+Pron+Rel+Pl+Nom ma
mij+Pron+Rel+Pl+Nom mahá

from lang-smj.

Comment 4393

Date: 2011-06-04 15:08:13 +0200
From: Trond Trosterud <<trond.trosterud>>

Here is the full mij poaradigm, with the restrictive generator (excluding +Use/NG:

~/main/gt$cat test.txt | lookup -flags mbTT smj/bin/ismj-SW.restr.fst
0%>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>100%
mij+Pron+Rel+Sg+Nom mij

mij+Pron+Rel+Sg+Gen man

mij+Pron+Rel+Sg+Acc majt

mij+Pron+Rel+Sg+Ill masi

mij+Pron+Rel+Sg+Ine manna

mij+Pron+Rel+Sg+Ela massta

mij+Pron+Rel+Sg+Com majna

mij+Pron+Rel+Ess manen

mij+Pron+Rel+Pl+Nom mahá

mij+Pron+Rel+Pl+Gen maj

mij+Pron+Rel+Pl+Acc mejt

mij+Pron+Rel+Pl+Ill mejda

mij+Pron+Rel+Pl+Ine majn

mij+Pron+Rel+Pl+Ela majsta

mij+Pron+Rel+Pl+Com maj

Case closed.

from lang-smj.