GiellaLT provides an infrastructure for rule-based language technology aimed at minority and indigenous languages, and streamlines building anything from keyboards to speech technology. Read more about Why. See also How to get started and our Privacy document.
Hente nye par:
Gå til mappa over apertium-xxx-yyy, som hos meg heiter apertium
git clone https://github.com/apertium/apertium-kaz-kir.git
med eit anna par enn kaz-kir viss du heller vil ha det (t.d. sme-deu, …)
Kv
for lang-sma, …:
./configure --enable-apertium ...
make
for sme-sma, sme-nob, and sme-smj:
git switch rtx
./autogen.sh --with-lang1=../lang-xxx/tools/mt/apertium --with-lang2=../lang-yyy/tools/mt/apertium
./configure --with-lang1=../lang-xxx/tools/mt/apertium --with-lang2=../lang-yyy/tools/mt/apertium
./configure --with-lang1=../lang-sme/tools/mt/apertium --with-lang2=../lang-sma/tools/mt/apertium
Concrete example, sme-smj
./configure --with-lang1=$GTLANGS/lang-sme/tools/mt/apertium --with-lang2=$GTLANGS/lang-smj/tools/mt/apertium
number = (ND sg) sg du pl ; ! if number is missing, use <ND> or <sg>
number > number : du pl ; ! replace all <du> with <pl>
1.gender/sl means the same as <clip pos="1" side="sl" part="gender"/>
prpers<prn><pers><p3><sg>/el<prn><pers><p3><sg>/Albert<np><m><sg>
SIDE_SOURCES = sl tl ; ! change default order of unspecified clips
SIDE_SOURCES = tl ; ! if you just want tl when unspecified (1.gender), and have to be explicit (1.gender/sl) when you want sl
NP -> det n { 1[number=2.number/sl] _ 2 } ;
NP -> adj n.number = <adj> <n><number> ! don't do this :)
NP -> adj n.sg = <adj> <n><sg>
NP -> adj n.$number = <adj> <n><sg/du/pl>
NP -> adj n [$number=2.number] { 1 _ 2 } ;
NP -> det adj n { 1 _ 2 _ 3 } ;
NP -> n { 1 } ;
sg_or_pl = sg pl ;
NP -> adj n.[sg_or_pl]
NP{ AP{ adv adj } n }
NP
/ \
AP n
/ \
adv adj
! define NP's tags
NP: <NP>.number.gender ;
! define NP's structure
NP -> adj n.$number.$gender { 1 _ 2[number=$number,gender=$gender] } ;
! when analyzing, find an <adj> then an <n>, combine them into an NP phrase, give this NP node the number and gender tags by taking them from the noun. When generating this NP, output the adjective, then a space, then the noun, and set the number tag of the noun to be the number tag of the NP and the same for gender (since these tags may have been changed by something elsewhere in the tree).
NP -> adj %n { 1 _ %2 } ;
! as above, but don't explicitly list the tags to be copied at each step
example: https://wiki.apertium.org/wiki/File:Recursive_Transfer_Hobbit_Example.pdf
VP -> %vblex DP { 1[tense=$tense, person=$person, number=$number] _ 2 } |
##%vblex DP= analyse
{ ... } = genrerert output
3 output:
## 1=%vblex
_ = melllomrom
## 2 = DP
## % = ta verdien
adv %VP (if (1.lem/sl = there) { %2 } else { 1 _ %2 } ) |
adv %VP { (if (1.lem/sl = there) [ 1 _ ]) %2 } |
PP %VP { 1 _ %2 } ;
eksempel:
PP -> pr DP { 1 _1 2 } ; (mellomrom i output) "i skogen"
NP -> n guio n { 1 2 3 } ; "skuvle-busse" guio = bindestrek
<b pos="1"/> = _1 #old way
Kevins rule: (ikke mellomrom i output)
showing recursiveness
a_cmp = cmp cmp-split ;
N -> %n ?(1.a_cmp = "") { %1 }; ! matches ^møte<n>$
N -> n.*.cmp %N { 1 %2 } ; ! matches ^kake<n><cmp>$ followed by N
! oste-kakemøter ^ost<n><sg><cmp-split>$^kake<n><sg><cmp>$^møte<n><pl>$
! analyses into this tree structure:
^møde<N><pl>{
^ost<n><sg><cmp-split>/ost<n><sg><cmp-split>$
^møde<N><pl>{
^kake<n><sg><cmp>/kage<n><sg><cmp>$
^møde<N><pl>{
^møte<n><pl>/møde<n><pl>$
}$
}$
}$
?(1.a_cmp = "")
example rule from Kevin’s file (nor-swe):
N -> "N→A" %n ?(1.mainpos=adj) { %1(A)[lemcase=$lemcase] } ;
! from fin-nob https://github.com/apertium/apertium-fin-nor/blob/master/apertium-fin-nor.fin-nob.rtx
!!!!!!!!!!!!!!!
!! TAG-REWRITE RULES
!!!!!!!!!!!!!!!
tense > tense : pri pres, past pii;
case > case : ine dat, ela dat, ill dat, ade dat, abl dat, all dat, abe dat, com dat, ins dat, lat nom, par acc, ess acc;
inf > inf : infa inf, infe inf, infma inf ;
!!!!!!!!!!!!!!!
!! REDUCTION RULES for negation
!!!!!!!!!!!!!!!
VP -> ei@vaux vblex { 2[number=1.number,person=1.person,tense=pri] _ ikke@adv } |
ei@vaux vaux { 2[number=1.number,person=1.person,tense=pri] _ ikke@adv } |
ei@vaux vbmod { 2[number=1.number,person=1.person,tense=pri] _ ikke@adv } ;
NEG -> ei@vaux { ikke@adv } ; ! ^ikke<adv>$ -> ikke
! from fin-eng https://github.com/apertium/apertium-fin-eng/blob/master/apertium-fin-eng.fin-eng.rtx
to run tests
apertium-regtest web -c 'rtx$'Examples:
apertium-regtest web -c '^sme-deu$'
echo 'Mun oainnán stuorra girjjiid viesu duohken.' | apertium -d . sme-nob_rtx-tree
Sjekk ut nye apertium-modular.
git clone https://github.com/apertium/apertium-sme-nob.git
gå hit i nettlesaren:
og sök
More examples:
make warning, what does it mean:
: Warning at line 9480, column 67: Paths to rule 125 blocked by rule 80.
since the exact same pattern applies for both rules, rule 125 will never apply
apertium-regtest web -p 3001 -c 'rtx$' # running on a different port, since one is running already
"borrat" <mv> V <PA-Loc-Food> <AG-Nom-Ani> <PA-Acc-Any> TV <AG-Nom-Ani>
echo 'Mánát borret gávpogis ja gilis' | apertium -d . sme-sma_rtx-tree # draw the analysis tree
echo 'Mánát borret gávpogis ja gilis' | apertium -d . sme-sma_rtx-transfer # output of transfer
echo "Mánát borret varas láibbiid." | apertium -d . sme-deu
die Kinder essen frischen Brote.
> Die Kinder essen frische Brote. (strong adj not mix)
du<s n="prn"/>
^don<prn><pers><p2><sg><gen><@→N>/du<prn><pers><p2><mf><sg><gen><@→N>/dein<det><pos><p2><mf><sg><gen><@→N>$
https://wiki.apertium.org/wiki/List_of_symbols how to connect this to that:
<e><p><l>Sámediggi<s n="np"/></l><r>Saamisches<b/>Parlament<s n="np"/></r></p></e> ---- strong
<e lm="saamisch"> <i>saamisch</i><par n="abenteuerlich__adj"/></e>
with article it should be weak
das Saamische Parlament
des Saamischen Parlamentes
Multiwords are manually written in an additional xml-format dictionary. (*.lsx)
<e><p><l>Sámediggi<s n="np"/></l><r>Saamisches<b/>Parlament<s n="np"/></r></p></e>
álgit - anfangen
álggán - ich fange X an. — X - optional NP
álggán dainna áššiin - ich fange mit der Sache an
<e><p><l>álgit</l><r>anfangen</r></p><par n="vblex_sep"/></e>
<?xml version="1.0" encoding="UTF-8"?>
<dictionary type="separable">
<alphabet>aAáÁăảạÆbBcCčdDÐeEéèễệfFgGhHiIíjJkKlLmMnNoOóöớpPqQrRsSšştTŢuUúüưvVwWxXyYzZ</alphabet>
<sdefs>
<sdef n="adj"/>
<sdef n="adv"/>
<sdef n="pr"/>
<sdef n="n"/>
<sdef n="v"/>
<sdef n="vblex"/>
<sdef n="vbser"/>
<sdef n="det"/>
<sdef n="dat"/>
</sdefs>
<pardefs>
<pardef n="word">
<e><i><w/><t/><d/></i></e>
<e></e>
</pardef>
<pardef n="words">
<e><par n="word"/><par n="word"/><par n="word"/></e>
</pardef>
</pardefs>
<section id="main" type="standard">
<!--
^fange/fangen<vblex><pri><p1><sg>/fangen<vblex><prs><p1><sg>/fangen<vblex><prs><p3><sg>$ ^x/*x$ ^an/an<pr><dat>/an<pr><acc>$^./.<sent>$ -->
<e>
<p><l>fangen</l><r>anfangen</r></p>
<i><s n="vblex"/><t/><d/></i>
<par n="words"/>
<p><l>an<s n="pr"/><s n="dat"/><d/></l><r></r></p>
</e>
</section>
</dictionary>
===
sma multiword (Maja):
<e><p><l>heahppášit<s n="vblex"/><s n="tv"/></l><r>skaamoem<b/>vedtedh<s n="vblex"/><s n="tv"/></r></p></e>
skaamoem% vedtedh+Use/MT:skaamoem% vedt BÅETEDH_TV_ePRET ; !
!åvtese% buektedh+Use/MT:åvtese% buekt BÅETEDH_TV ;
==> thus this procedure cannot put anything between skaamoem and V.
====
<e><p><l>váldit<s n="vblex"/><s n="tv"/></l><r>sebrudahttet<s n="vblex"/><s n="tv"/></r></p></e> <!-- instead of váldit mielde (see lrx) -->
mij lip aj ávon gå li nåv moattes gudi sebrudahtti juojggusav ådåájggásasj musihkan > mii leat maiddái ilus go leat nu máŋggas geat váldet juoigama mielde ođđaáigásaš musihkkii
https://github.com/apertium/apertium-kaz-kir/blob/rtx/apertium-kaz-kir.kaz-kir.rtx
https://github.com/apertium/apertium-sme-deu/commit/08e486fdfc92a69ee170dd7bcc71d7f177ac2180
<!-- Examples from nob→nno for inspiration -->
<e lm="gi vår tilslutning til -> slutte oss til" c="slutte oss til">
<p><l>gi</l><r>slutte</r></p> <i><t/><d/></i>
<p><l>vår<t/><d/></l><r></r></p>
<p><l>tilslutning<t/><d/></l> <r></r></p>
<p><l>til<t/><d/></l><r></r></p>
<p><l></l><r>vi<s n="prn"/><s n="pers"/><s n="p1"/><s n="un"/><s n="sp"/><s n="acc"/><d/></r></p>
<p><l></l><r>til<s n="pr"/><d/></r></p>
</e>
<e lm="den enkelte" c="kvar enkelt">
<p>
<l>den<s n="det"/><s n="dem"/></l>
<r>hver<s n="det"/><s n="qnt"/></r>
</p>
<i><t/><t/><d/></i>
<p>
<l>enkelt<s n="adj"/><s n="pst"/><s n="un"/><s n="sp"/><s n="def"/><d/></l>
<r>enkelt<s n="adj"/><s n="pst"/><s n="nt"/><s n="sg"/><s n="ind"/><d/></r>
</p>
<par n="indefinte-noun-phrase"/>
</e>
<e lm="2. Petersbrev -> Andre Petersbrevet" c="Andre Petersbrevet">
<par n="1tilførste"/>
<par n="bibelbøker"/>
</e>
<e lm="ubefestet by -> by uten forsvarsanlegg" c="by utan forsvarsanlegg">
<p><l>ubefestet<t/><d/></l> <r></r></p>
<i>by<t/><d/></i>
<p><l></l> <r>uten<s n="pr"/><d/></r></p>
<p><l></l> <r>forsvarsanlegg<s n="n"/><s n="nt"/><s n="sg"/><s n="ind"/><d/></r></p>
</e>
<e lm="tilbringe + subst + med å -> bruke + subst + på å" c="bruke tid på å lese">
<p><l>tilbringe</l> <r>bruke</r></p> <i><t/><d/></i>
<par n="ANYWORD|"/>
<par n="ANYWORD|"/>
<i><w/><s n="n"/><t/><d/></i>
<p><l>med</l> <r>på</r></p> <i><t/><d/></i>
<i>å<t/><d/></i>
</e>
suomastit, sámastit, darostit
finnisch/saamisch/norwegisch sprechen
<e><!-- suomastit</l><r>Finnisch<b/>sprechen</r>-->
<p><l>sprechen</l><r>finnisch<b/>sprechen</r></p>
<i><s n="vblex"/><t/><d/></i>
<par n="words"/>
<p><l>finnisch<s n="adj"/><s n="sint"/><s n="pred"/><d/></l><r></r></p>
</e>
^váldit<vblex><tv><der_passl><vblex><iv><inf><@-FMAINV>$
^váldit<vblex><tv><pass><inf>...$
váldojuvvot
váldojuvvojit váldit+V+TV+Der/PassL+V+IV+Ind+Prs+Pl3
sie werden genommen
nehmen -> personal pronoun (person number agreement) + “werden” person number tense (from Sámi verb) + nehmen past participle
<e r="LR"><!-- PASSIVE VERBS -->
<i><w/></i>
<p><l><s n="vblex"/><s n="tv"/><s n="der_passl"/><s n="vblex"/><s n="iv"/></l><r><s n="vblex"/><s n="tv"/><s n="pass"/></r></p>
<i><t/><d/></i>
</e>
vblex: (if (1.pass = pass)
[ 1(vblex_simple)[pos_tag=vaux,lem=werden] _ 1(vblex_pp)[pcp=pp] ]
else 1(vblex_simple)) ;
bargagoahtá bargat+V+TV+Der/InchL+V+Ind+Prs+Sg3
jobbe
begynne å jobbe
vblex_simple: (if (1.finity = inf) 1(vblex_inf)
elif (1.finity = actio) 1(vblex_inf)[inf=inf] ! probably not ideal
elif (1.pcp = pp) 1(vblex_pp)
elif (1.pcp = prfprc) 1(vblex_pp) ! probably not ideal
elif (1.pcp = pprs) 1(vblex_pp)
else 1(vblex_fin) ) ;
!vblex: _.tense ;
vblex: (if (1.derp = der_passl)
[ 1(vblex_simple)[pos_tag=vblex,lem=bli] _ 1(vblex_pp)[pcp=pp] ]
elif (1.derp = der_inchl)
[ 1(vblex_simple)[lem=begynne] _ å@part _ 1(vblex_simple)[finity=inf,inf=inf] ]
else 1(vblex_simple)) ;
VP -> vblex { 1 } ;
Example (sme-smj):
váldojuvvojit váldit+V+TV+Der/PassL+V+IV+Ind+Prs+Pl3
sie werden genommen
nehmen -> personal pronoun (person number agreement) + “werden” person number tense (from Sámi verb) + nehmen past participle
<e r="LR"><!-- PASSIVE VERBS -->
<i><w/></i>
<p><l><s n="vblex"/><s n="tv"/><s n="der_passl"/><s n="vblex"/><s n="iv"/></l><r><s n="vblex"/><s n="tv"/><s n="pass"/></r></p>
<i><t/><d/></i>
</e>
vblex: (if (1.pass = pass)
[ 1(vblex_simple)[pos_tag=vaux,lem=werden] _ 1(vblex_pp)[pcp=pp] ]
else 1(vblex_simple)) ;
bargagoahtá bargat+V+TV+Der/InchL+V+Ind+Prs+Sg3
jobbe
begynne å jobbe
vblex_simple: (if (1.finity = inf) 1(vblex_inf)
elif (1.finity = actio) 1(vblex_inf)[inf=inf] ! probably not ideal
elif (1.pcp = pp) 1(vblex_pp)
elif (1.pcp = prfprc) 1(vblex_pp) ! probably not ideal
elif (1.pcp = pprs) 1(vblex_pp)
else 1(vblex_fin) ) ;
!vblex: _.tense ;
vblex: (if (1.derp = der_passl)
[ 1(vblex_simple)[pos_tag=vblex,lem=bli] _ 1(vblex_pp)[pcp=pp] ]
elif (1.derp = der_inchl)
[ 1(vblex_simple)[lem=begynne] _ å@part _ 1(vblex_simple)[finity=inf,inf=inf] ]
else 1(vblex_simple)) ;
VP -> vblex { 1 } ;
Example (sme-smj):
Danielgo borrá láibbi?
Don bat leat/bohtet?
PLAN
https://wiki.apertium.org/wiki/North_Saami_and_South_Saami/Regression_tests
https://wiki.apertium.org/wiki/North_Saami_and_Lule_Saami/Regression_tests
daenbeijjien:daen-bieijjien adv ; ! Dir/LR <- Apertium way
daenbeijjien adv ;
daenbejjien+Use/NG:daen-bieijjien adv ;
Ich erschreckte mich als ich das Auto sah.
Ich wundere mich warum Børre das gemacht hat.
Sie freut sich.
I want.person.number prn.refl.person.number negation-adv mainv.inf
Habe ich mich gefreut?
Ich setze mich.
Ich freue mich nicht.
LEFT: ^[lemma]<vblex><TAGS><p3><sg>$ ^sich<prn><ref><p3><mf><sp><nom>$
->
RIGHT: ^sich [lemma]<vblex><TAGS><p3><sg>$
<t/> match all tags given
<w/> match any letter given
<i><t/></i> output any other tags unchanged
<i> identity
<r></r> r is empty, so deleted
LEFT: ^*<vbmod|vblex><TAGS><p3><sg>$ ^sich<prn><...>$ ^[lemma]<vblex><inf>$
RIGHT: ^*<vbmod|vblex><TAGS><p3><sg>$ ^sich [lemma]<vblex><inf>$
<i><w/><s n="vbmod"/><t/><s n="p3"/><s n="sg"/><d/></i>
echo Don it ferte illudit | apertium -d. sme-deu-tree
Don it háliit oađđit -> Du möchtest nicht _schlafen.
PRN ii@vblex vblex vblex.*.inf { 1_ 3[person=2.person,number=2.number] _nicht@adv_ 4 }|
PLAN
split verb - auf, aus, …
<e><!-- -->
<p><l>geben</l><r>aufgeben</r></p>
<i><s n="vblex"/><t/><d/></i>
<par n="words"/>
<p><l>auf<s n="pr"/><s n="dat"/><d/></l><r></r></p>
</e>
<pardef n="auf_verb">
<e><i>geben</i></e>
</pardef>
<e><!-- -->
<p><l></l><r>auf</r></p>
<par n="auf_verb"/>
<i><s n="vblex"/><t/><d/></i>
<par n="words"/>
<p><l>auf<s n="pr"/><s n="dat"/><d/></l><r></r></p>
</e>
NP -> num n [$number=(if (1.lem = "1" or 1.lem = "okta") sg else pl)] { 1 _ 2 } ;
$ 1 2
$number, 1.number, 2.number
! number and noun will become plural ( no use for this)
NP -> num n [$number=(if (1.lem = "1") sg else pl)] { %1 _%2 } ;
! number and noun will remain singular (sme -> smj)
NP -> num n [$number=(if (1.lem = "1") sg else pl)] { 1_ 2 } ;
! noun will become plural but number will remain singular (sme -> nob)
NP -> num n [$number=(if (1.lem = "1") sg else pl)] { 1 _%2 } ;
! number stays the same, noun gets same number tag as number, as long as the num is sg
NP -> num n [$number=(if (1.lem = "1") sg else pl)] { 1_ 2[number=1.number] } ;
NP -> %np { %1 } ;
NP -> %np { 1 } ;
{ %1 } ! we want to change tags on this based on something elsewhere in the tree
{ 1 } ! the tags are probably already correct from the bidix, so just keep the input tags
rule ordering: either more specific first or put a weight
VP -> vblex { 1 } ;
VP -> 2: vblex.*.neg { 1 } ; #verb phrase could be a negative verb, pick the higher weight
habitive structure:
@ADVLhab→ (@ADVL> <hab>)
@←SUBJext
Beavddis @ADVL-ine→ lea dihtor @←SUBJext.
Loc @ADVL-ine→ @ADVL-ela→ Adv @ADVLcs-ela→
NP -> %n { %1[case=(if (1.case = loc and (1.function = "@←ADVL-ine" or 1.function = "@→ADVL-ine")) ine
if (1.case = loc and (1.function = "@←ADVL-ela" or 1.function = "@→ADVL-ela")) ela
else 1.case)] } ;
so:
function = "@→SUBJ" "@←OBJ" "@→ADVL" "@→ADVL-ine" "@→ADVL-ela" ;
NP: _.function;
S -> NP."@→SUBJ" VP |
VP NP."@←SUBJ"
NP: _.number ;
VP:_.number ;
S: _;
NP -> %n { 1 } ;
! send number from input verb to VP, by default get it back again the same, unless the next phrase up changes it
VP -> %vblex { %1 } ;
! if NP and VP disagree in number, change VP
S -> NP VP { 1_ 2[number=1.number] } ;
Analysis:
S
/ \
NP VP
| |
n vblex
n has number from input
vblex has number from input
NP has number from n
VP has number from vblex
S doesn't have number
Generation
NP VP
| |
n vblex
NP has the number it had before
VP now has the number from NP
NP -> %n { 1[poss=$poss] } ;
NP: _.number.case.poss ;
NP -> %n { %1 } ;
same as
NP -> %n { 1[number=$number,case=$case,poss=$poss] } ;
NP -> %n { %1[number=sg] } ;
same as
NP -> %n { 1[number=sg,case=$case,poss=$poss] } ;
NP -> 2: %n.coll [$number=pl] { 1 } ;
! if we have a collective numeral, apply this rule rather than the normal NP -> n rule (weight 2 so it wins) when copying tags to the phrase, don't copy the morphological <sg> but rather always put <pl> so that we don't screw up verb agreement, and don't put % inside the {} so that we keep the input tags and don't override the <sg> tag on the noun
váruhit
sme-deu:
apertium -d . sme-deu
Loc > auf + X (Acc)
echo mun váruhan áššis. | apertium -d . sme-deu
ich passe in der Sache auf#.
Dealing with NP changes (case, prepositions) based on verb valency:
VP -> %VP NP.*.ill { %1 _ 2[case=(if (1.lem in_caseless verb_ill_to_ine) ine
else 2.case)]
_ %1};
<TH-Loc-Any>
verb_loc_to_auf = váruhit ;
VP -> %VP NP.*.loc { %1 _ 2[case=(if (1.lem in_caseless loc_to_auf)
else 2.case)]
_ %1};
Theme = "Th-Loc-Any" "Th-Loc-Ine" ;
VP: _.Theme ;
VP -> %
case = nom acc loc pr_auf ;
NP -> %n (if ($case = pr_auf) { auf@pr.acc _%1[case=acc] }
else { %1 }) ;
VP -> %VP NP.*.loc { %1_ 2[case=(if (1.lem in_caseless verb_loc_to_auf) pr_auf
else 2.case)] } ;
∈cl = in_caseless
=
equal
https://wiki.apertium.org/wiki/North_Saami_and_Lule_Saami/Regression_tests
https://wiki.apertium.org/wiki/North_Saami_and_South_Saami/Regression_tests
lea vejolaš
lea go x vejolaš
(sme) Daid galget almmolaš eiseválddit geavahit, oktan vejolaš dárogiel namain. → Dejtie byjjes åajvaladtjh gelkieh nuhtjedh, nuepies daaroengïeleh nommine ektine.
Ich gewöhne mich in eine #Schule
change noun based on the verb, so we want to attach it to the VP so write a rule to attach an illative NP
| search for the last VP rule and add a | and a new condition |
copy lemma of the verb to the phrase so we can access it for the valency
we need change from illative to a prepositional phrase with “an”
adding a weight “2:” because there is a conflicting rule change illative case if the lemma is in the set verb_ill_to_an
add a new case to the set of case the NP gets a new case tag from the rule “pr_an”
case = (CD nom) nom acc dat gen ill loc com ess CD pr_auf_acc pr_auf_dat pr_an_acc ;
verb_ill_to_an = "sich gewöhnen" ;
VP -> 2: %VP NP.*.ill { 1 _ 2[case=(if (1.lem in_cl verb_ill_to_an) pr_an_acc else 2.case)] } ;
NP -> 2: %N (if ($case = pr_auf_acc) { auf@pr.acc _ %1[case=acc] }
! assume that we came from a PP and should thus be definite
if ($case = pr_an_acc) { an@pr _ 1[maybe_art](def=def) %1[case=acc] }
else { %1 }) ;
VP -> %lea@VP vejolaš@adj.*.nom { %1[lem=2.lem/tl] } ;
son lea vihtta jagi boaris. > Sån la vidá jagák
coordination
! agree with noun nearest the verb (if the verb comes after)
NP -> %NP cnjcoo NP [$number=3.number] { 1 _2_ 3 } ;
! number is dual if both conjuncts are human (translating to Sámi)
conj_dual = sem_hum sem_org sem_fem sem_mal sem_sur;
NP -> %NP cnjcoo NP
[$number=(if (1.number = sg and 3.number = sg and
1.sem in conj_dual and 3.sem in conj_dual) du
else pl)]
{ 1 _2_ 3 } ;
Ánde og Máret spiser. -> borraba
cat article.txt | apertium -d . sme-smj_rtx > translated_article.txt