130828

Apertium sme-sma 28.8.

Present: Francis, Lene, Trond.

Agenda:

  • Tag reording
  • Other issues

Tag reordering

bidix                 fst                           result
---------------------------------------------------------------
skuvla<n> .INTERSECT. skuvla<sem_org><sem_plc><n> = ()
skuvla<n> .INTERSECT. skuvla<n><sem_org><sem_plc> = (skuvla<n>)
skuvla<n> .INTERSECT. skuvla<n>                   = (skuvla<n>)
---------------------------------------------------------------

But:

<e><p><l>Stockholm<s n="n"/><s n="prop"/><s n="plc"/></l><r>Stuehkie<s n="n"/><s n="prop"/></r></p></e>

solutions:

  1. move tags
  2. delete tags

So:

We do the following

  1. Reshuffle tags in fst:
    1. prop tags are as they are from fst
    2. noun are reordered by a script in sme/src/filters
    3. adverbs are as they are from fst
  2. the tags are then matched for the pruning of the fst in apertium
  3. The bidix does not contain semtags
  4. The cg in Apertium removes the tags

here is the explicit description of what we want: 

---------------------------------------------------------------
fst:
<n><sem_org><sem_plc><sg>...     after reshuffle
<n><prop><sem_plc><sg>...        from the fst
<adv><sem_time>

cg: semtags in, nosemtags out

bidix: no sem-tags
---------------------------------------------------------------

Other issues

jagi 1974

Son lea riegádan "jagi 1974" ja bajásšaddan "Romssas".
Dihte "jaepien 1974" reakedi jih "Tromsene" byjjesovvi.

^Son/Son<pron><pers><sg3><nom><@SUBJ→>$ ^lea/leat<v><iv><ind><prs><sg3><@+FAUXV>$ ^riegádan/riegádit<v><iv><prfprc><@-FMAINV>$ ^jagi/jahki<n><sg><gen><@←ADVL>$ ^1974/1974<num><sg><nom><@N←>$ ^ja/ja<cc><@CNP>$ ^bajásšaddan/bajásšaddat<v><iv><prfprc><@-FMAINV>$ ^Romssas/Romsa<n><prop><sg><loc><@←ADVL>$^../..<clb>$

$ echo "Son lea riegádan jagi 1974 ja bajásšaddan Romssas." | apertium -d . sme-sma
Dïhte jaepien 1974 reakadamme jïh Tromsøesne byjjenamme.

^pron-pers<SN><@SUBJ→><PX>{^Dïhte<pron><pers><sg3><nom>$}$ ^v<SV><@+FMAINV>{^reakadidh<v><iv><prfprc>$}$ ^n-num<SN><@←ADVL><PX>{^jaepie<n><sg><gen>$ ^1974<num><sg><nom>$}$ ^cc<CC><@CNP>{^jïh<cc>$}$ ^prfprc<SV><@-FMAINV>{^byjjenidh<v><iv><prfprc>$}$ ^n<SN><@←ADVL><PX>{^Tromsø<n><prop><sg><ine>$}$^sent<SENT><@X>{^..<clb>$}$

"<Son>"
    "son" Pron Pers Sg3 Nom @SUBJ> 
"<lea>"
    "leat" V IV Ind Prs Sg3 @+FAUXV 
"<riegádan>"
    "riegádit" V IV PrfPrc @-FMAINV 
"<jagi>"
    "jahki" N Sg Gen @<ADVL 
"<1974>"
    "1974" Num Sg Nom @N< 
"<ja>"
    "ja" CC @CNP <====================  @CVP would tell that there is coming a new finite verb
"<bajásšaddan>"
    "bajásšaddat" V IV PrfPrc @-FMAINV 
"<Romssas>"
    "Romsa" N Prop Sg Loc @<ADVL 
"<.>"
    "." CLB

Links: 

Before:
$ echo "Nieiddat leat čeahpit. " | apertium -d . sme-sma
Nïejth leah væjkelh. 

After:
$ echo "Nieiddat leat čeahpit. " | apertium -d . sme-sma
Nïejth leah væjkele. 

     ^buot/buot<pron><indef><attr><@→N>$ ^gielaid/giella<n><pl><acc><@←OBJ>$
     ^attr-n<SN><@←OBJ><PX>{^gaajhke<pron><indef><pl><acc>$ ^gïele<n><pl><acc>$}$
     gaajhkide gïelide

echo "Mun boađán boahtte jagi." | apertium -d . sme-sma
Manne båetije jaepien båatam.

båetije båetedh+V+IV+PrsPrc
båetije båetedh+V+IV+Der/NomAg+N+Sg+Nom

Here is a solution

<e><p><l>boahtte<s n="a"/></l><r>båetije<s n="a"/></r></p></e>

  • ND = number to be determined (take it from the noun)
  • CD = case to be determined (take it from the noun)

Lexicon entry form: 

    
    <e r="LR"><p><l>boahtte<s n="a"/><s n="attr"/></l><r>båetije<s n="a"/><s n="ND"/><s n="CD"/></r></p></e>
    <e r="RL"><p><l>boahtte<s n="a"/><s n="attr"/></l><r>båetije<s n="a"/></r></p></e>

Now: 

^pron-pers<SN><@SUBJ→><PX>{^Manne<pron><pers><sg1><nom>$}$ ^v<SV><@+FMAINV>{^båetedh<v><iv><ind><prs><sg1>$}$ ^a-n<SN><@←ADVL><PX>{^båetije<a><sg><gen>$ ^jaepie<n><sg><gen>$}$^sent<SENT><@X>{^..<clb>$}$

$ echo "Mun boađán boahtte jagi." | apertium -d . sme-sma
Manne #båetije jaepien båatam.

So, what is the role of båetijen? 

boahtte jagis -> båetijen jaepesne
attr     loc     gen.attr      ine

båetije båetije+A+Attr <=== remove Attr and give cases
båetijen    båetije+A+Gen+Attr

båetije båetije+A+Sg+Nom 
båetijen båetije+A+Sg+Acc Use/MT
båetijen båetije+A+Sg+Gen 
båetijen båetije+A+Sg+Ine 
båetijen båetije+A+Sg+Ela 
båetijen båetije+A+Sg+... 


sme:
buori   buorre+A+Sg+Gen
buori   buorre+A+Sg+Acc

buorre
buorre  buorre+A+Sg+Nom

boahtte boahtte+A+Attr => båetije/båetijen

Pron+Dem       @→N = +Det+Dem (?)
Pron+Dem +Attr @→N = +Det+Dem (?)

We then have four types:

indecl sme -> decl sma

    <e r="LR"><p><l>boahtte<s n="a"/><s n="attr"/></l><r>båetije<s n="a"/><s n="ND"/><s n="CD"/></r></p></e>
    <e r="RL"><p><l>boahtte<s n="a"/><s n="attr"/></l><r>båetije<s n="a"/></r></p></e>

indecl sme -> indecl sma

-

decl sme -> decl sma

-

decl sme -> indecl sma

    <e r="LR"><p><l>dehálaš<s n="a"/></l><r>vihkeles<s n="a"/><s n="indecl"/></r></p></e>
    <e r="RL"><p><l>dehálaš<s n="a"/></l><r>vihkeles<s n="a"/></r></p></e>
  • vihkeles+A+Superl+Sg+Nom vihkelommes
  • vihkeles+A+Sg+Acc vihkelem

He sees the X cat. 

in langs/sma:
./configure --with-hfst --enable-apertium --enable-oahpa
the file is 
sma/src/morphology/*/adjectives-oahpa.lexc

Lexical selection

boahtit oidnosii -> våajnoes sjïdtedh

Default translation for boahtit is båetedh, alternative translation is sjïdtedh 

    <e slr="1"><p><l>boahtit<s n="v"/><s n="iv"/></l><r>båetedh<s n="v"/><s n="iv"/></r></p></e>
    <e slr="2"><p><l>boahtit<s n="v"/><s n="iv"/></l><r>sjïdtedh<s n="v"/><s n="iv"/></r></p></e>

apertium-sme-sma.sme-sma.lrx 

  <rule>
    <match lemma="boahtit" tags="v.*"><select lemma="båetedh" tags="v.*"/></match>
  </rule>

  <rule>
    <match lemma="boahtit" tags="v.*"><select lemma="sjïdtedh" tags="v.*"/></match>
    <match lemma="oidnosii"/>
  </rule>

^boahtit<v><iv><ind><prs><pl3><@+FMAINV>/båetedh<v><iv><ind><prs><pl3><@+FMAINV>/sjïdtedh<v><iv><ind><prs><pl3><@+FMAINV>$ ^oidnosii<adv><@←ADVL>/våajnoes<adv><@←ADVL>$

^Dat/dat<pron><dem><pl><nom>/dat<pron><dem><sg><nom>$ ^boahtá/boahtit<v><iv><ind><prs><sg3>$ ^oidnosii/oidnosii<adv>$^./.<clb>$

  1. Artihkele 7: m *geatnegahttá nasjonaalestaatide tjïrrehtidh konkreetide råajvarimmide vaarjeleminie åvteste unnebelåhkoengïeli nimhtie ahte dah våajnoes sjidtieh dovne politihkesne, laakine jïh åtnosne.

^pron-dem<SN><@SUBJ→><PX>{^dïhte<pron><dem><pl><nom>$}$ ^v<SV><@+FMAINV>{^sjïdtedh<v><iv><ind><prs><pl3>$}$ ^adv<ADV><@←ADVL>{^våajnoes<adv>$}$

<ADV><@←ADVL>

$ bash dev/test-prefix.sh "juoga<pron><indef>"

error test report

sh generation-errors.sh | grep '#' | grep '<pron'
10   ^jïjnje<pron><indef><pl><pl><nom>$ #jïjnje\<pron\>\<indef\>\<pl\>\<pl\>\<nom\> ^jïjnje/jïjnje<a><attr>/jïjnje<a><sg><nom>/jïjnje<adv>/jïjnje<pron><indef><sg><nom>$^./.<clb>$
2    ^måedtie<pron><indef><sg><ine>$    #måedtie\<pron\>\<indef\>\<sg\>\<ine\>  ^måedtie/måedtie<pron><indef><sg><nom>$^./.<clb>$
2    ^aktege<pron><indef>$  #aktege\<pron\>\<indef\>    ^aktege/akte<num><sg><nom><foc_ge>/aktege<pron><indef><attr>$^./.<clb>$
1    ^naan<pron><indef><sg><nom>$   #naan\<pron\>\<indef\>\<sg\>\<nom\> ^naan/naan<pron><indef>$^./.<clb>$
1    ^naan<pron><indef><pl><acc>$   #naan\<pron\>\<indef\>\<pl\>\<acc\> ^naan/naan<pron><indef>$^./.<clb>$
1    ^dagkeres<pron><dem><pl><com><attr>$   #dagkeres\<pron\>\<dem\>\<pl\>\<com\>\<attr\>   ^dagkeres/dagkeres<a><attr>/dagkeres<pron><dem><attr>$^./.<clb>$

SVO - SOV with adjectives

SVO to SOV, but also S V AN to S AN V

  • Son oaidná skuvlla. Son oaidná ođđa skuvlla.
    • Before the fix: Satne skuvlem vuajna. Satne vuajna orre skuvlem.
    • After the fix: Son oaidná ođđa skuvlla. Dïhte orre skuvlem vuajna. 
^pron-pers<SN><@SUBJ→><PX>{^Dïhte<pron><pers><sg3><nom>$}$     <rule comment="REGLA: pron-pers">
^v<SV><@+FMAINV>{^vuejnedh<v><tv><ind><prs><sg3>$}$            <rule comment="REGLA: verb-fin"> 
^a-n<SN><@←OBJ><PX>{^orre<a><attr>$ ^skuvle<n><sg><acc>$}$     <rule comment="REGLA: adj-rdep n">
^sent<SENT><@X>{^..<clb>$}$

The trick was to define a phrase (SN) in the t1x file, and then refer to that phrase later.