Language Technology at UiT The Arctic University of Norway

The Divvun and Giellatekno teams build language technology aimed at minority and indigenous languages

View GiellaLT on GitHub divvungiellatekno/giellalt.uit.no

Page Content

Møte 16. desember 2016

Hangouts: Kevin, Trond, Lene

Saker:

  1. Strukturen i sme-nob vs. sme-smX
  2. Debugging av regr test
  3. Modes
  4. hfst vs bin (notater)
  5. Feilmeldingar
  6. Regresjonstestar

Strukturen i sme-nob vs. sme-smX

hfst vs bin

Kommando for å analysere sme:

echo ja|hfst-lookup .deps/sme-nob.automorf-trimmed.hfst
echo ja|hfst-lookup .deps/sme.automorf.hfst

make nob-sme # lagar nob→sme-filene
echo "sátni<n><sg><acc>" | hfst-lookup .deps/sme.autogen.hfst

Norsk skal vi analysere slik:

echo "hei" |  lt-proc -e ../../languages/apertium-nob/nob.automorf.bin
(-e må med for å ha analyse av (dynamisk) samansette ord)

$ echo "hei" | apertium -d ../../languages/apertium-nob nob-morph
^hei/heie<vblex><imp>/hei<n><m><sg><ind>/hei<n><f><sg><ind>/hei<n><m><sg><ind>/hei<n><f><sg><ind>$^./.<sent><clb>$

Slik gjer Kevin (denne klarer også dynamisk samansette ord):
$ echo "hei" | apertium -d . unob-sme-morph
echo Regjeringen | apertium -d . unob-sme-morph
^Regjeringen/Regjeringen<np><cog>/regjering<n><m><sg><def>/regjering<n><m><sg><def>$^./.<sent><clb>$

"<Ráđđehus>"
        "ráđđehus" N <sme> Sem/Org Sg Nom


 <e><p><l>ráđđehus<s n="n"/><s n="org"/></l><r>regjering<s n="n"/></r></p><par n="m_RL_f__n"/></e>



<e><p><l>ráđđehus<s n="n"/><s n="sem_org"/></l><r>regjering<s n="n"/></r></p><par n="m_RL_f__n"/></e>
<e><p><l>Regjeringen<s n="np"/><s n="sem_org"/></l><r>Regjeringen<s n="np"/></r></p></e>

nob-morph skil seg frå lt-proc ved å godta reserverte teikn.

(og hugsar alltid å ta med -e)

Debugging av regr test

Pronomen

echo jeg|apertium -d nob nob-morph
^jeg/jeg<n><nt><sg><ind>/jeg<n><nt><pl><ind>/jeg<prn><pers><p1><mf><sg><nom>$^./.<sent><clb>$

$ echo "mun lean dáppe"|apertium -d. sme-nob-postchunk
^jeg<prn><pers><p1><mf><sg><nom>$ ^være<vblex><pres>$ ^her<adv>$^.<sent><clb>$

echo "mun lean dáppe"|apertium -d. sme-nob
#jeg er her

$ echo '^jeg<prn><pers><p1><mf><sg><nom>$'|lt-proc -g sme-nob.autogen.bin
jeg

echo '^jeg<prn><pers><p1><mf><sg><nom>$'|lt-proc -g sme-nob.autogen.bin
#jeg

apertium-sme-nob$ echo "mun lean dáppe"|apertium -d. sme-nob-dgen
#jeg<prn><p1><mf><sg><nom> er her

echo "Son oaidná mu."|apertium -d. sme-nob-postchunk
^Han<prn><pers><p3><m><sg><nom>$ ^se<vblex><pres>$ ^jeg<prn><pers><p1><mf><sg><acc>$^..<sent><clb>$

cd ../../languages/apertium-nob svn up && make

Taggstrengane er identisk (gjeld også 3. person), men vi får likevel #.

Eigennamn

<e><p><l>bidjat<s n="vblex"/><s n="tv"/></l><r>sette<s n="vblex"/><s n="pers"/></r></p><par n="__verb"/></e>

Mathisen sem_sur - Mathisen cog

<e><p><l>Mathisen<s n="np"/></l><r>Mathisen<s n="np"/></r></p><par n="__np"/></e>
    <pardef n="__np" c="The nob tags are not output by transfer.">

$ echo Mathisenin | apertium -d . sme-nob-dgen
som #Mathisen<np>

^Mathisen<np><cog><ess><@HNOUN>$^.<sent>$

echo Mathisenin | apertium -d. sme-nob
som Mathisen

echo don leat Mathisenin | apertium -d. sme-nob
#du er Mathisen

 echo "Son oaidná mu."|apertium -d. sme-nob-dgen
Montro ser #jeg<prn><p1><mf><sg><acc>.

tf4-hsl-m0024:apertium-sme-nob trond$ echo "Mun lean dáppe."|apertium -d. sme-nob-dgen
#Jeg<prn><p1><mf><sg><nom> er her.

tf4-hsl-m0024:apertium-sme-nob trond$ echo "Son oaidná mu."|apertium -d. sme-nob-morph
^Son/son<pcle>/son<prn><pers><p3><sg><nom>$ ^oaidná/oaidnit<vblex><tv><indic><pres><p3><sg>$ ^mu/mun<prn><pers><p1><sg><acc>/mun<prn><pers><p1><sg><gen>$^../..<sent>$
Lene:
echo "Son oaidná mu."|apertium -d. sme-nob
Han ser meg.

sme-nob          Maŋŋá go Máhttolokten doaibmagođii, de lea dát ruhtadoarjja jávkan.
        - Etter at Kunnskapsløftet begynte å fungere, så har denne pengestønaden forsvunnet.
        + Etter at Kunnskapsløftet begynteåfungere, så har denne pengestønaden forsvunnet.

(ev. endra sem_sur→cog i bidix; sidan me matchar på taggane etter “cog”-taggen òg)


apertium-sme-nob$ echo Várggát | apertium -d. sme-nob
Vardø
apertium-sme-nob$ echo Mun lean Várggáin | apertium -d. sme-nob
#Jeg er i Vardø

apertium-sme-nob$ echo Mun lean Várggáin | apertium -d. sme-nob-syntax
"<Mun>"
        "mun" prn pers p1 sg nom @SUBJ→ MAP:1673:subj>Pers
"<lean>"
        "leat" vblex iv indic pres p1 sg @+FMAINV
"<Várggáin>"
        "Várggát" np top pl loc @←ADVL-ine MAP:1865:V<advl SELECT:2249
* **"Várggát" np pl loc @←ADVL-ine MAP:1865:V<advl SELECT**: 2249
"<.>"
        "." sent

apertium-transfer: Rule 46 Mun<prn><pers><p1><sg><nom><@SUBJ^Prn<SN><@SUBJ→><p1><mf><sg><nom>{^jeg<prn><pers><p1><mf><sg><nom>$}$ ^vcop<SV><@+FMAINV><indic><pres><p1><sg><impers><NC>{^være<vblex><pres>$}$ ^caseprep<PR><loc>{^i<pr>$}$ ^nom<SN><ind><pl><loc><impers>{^Vardø<np><top>$}$^default<default>{^.<sent><clb>$}$

echo Mun lean Lene Antonsen | apertium -d. sme-nob-chunker

apertium-transfer: Rule 46 Mun<prn><pers><p1><sg><nom><@SUBJ^Prn<SN><@SUBJ→><p1><mf><sg><nom>{^jeg<prn><pers><p1><mf><sg><nom>$}$ ^vcop<SV><@+FMAINV><indic><pres><p1><sg><impers><NC>{^være<vblex><pres>$}$ ^pre_nom<SN><@←SPRED><ind><f><nom><pers>{^Lene<np><f>$ ^Antonsen<np>$}$^default<default>{^.<sent><clb>$}$

apertium-sme-nob$ usme
Ammerud
Ammerud        Ammerud+N+Prop+Sem/Sur+Sg+Nom
Ammerud        Ammerud+N+Prop+Sem/Sur+Sg+Gen
Ammerud        Ammerud+N+Prop+Sem/Sur+Sg+Acc
Ammerud        Ammerud+N+Prop+Sem/Plc+Sg+Nom
Ammerud        Ammerud+N+Prop+Sem/Plc+Sg+Gen
Ammerud        Ammerud+N+Prop+Sem/Plc+Sg+Acc

22 jahkásaš Tine Chris Mathisen fárrii áibbas okto máddin Sápmái.
#22<det><qnt>un<pl> år gamle Tine Chris Mathisen flyttet helt alene sørfra til Sameland.

$ echo 22 jahkásaš Tine Chris Mathisen fárrii áibbas okto máddin Sápmái.|apertium -d . sme-nob-dgen

$ echo '^Tine<np><ant><f>$'|lt-proc -g ../../languages/apertium-nob/nob.autogen.bin
Tine

^22<num><arab><sg><nom>$ ^ihásâš<adj>$ ^Tine<np><ant><f><attr>$ ^Chris<np><ant><m><attr>$ ^Mathisen<np><sg><nom>$ ^varriđ<vblex><indic><pret><p3><sg>$ ^aaibâs<adv>$ ^ohtuu<adv>$ ^mäddi<n><ess>$ ^Säämi<np><top><sg><ill>$^.<sent>$^.<sent>$

^22<det><qnt>un<pl>$ ^år gammel<adj><sint><pst><un><pl><ind>$ ^Tine<np><f>$ ^Chris<np><m>$ ^Mathisen<np>$ ^flytte<vblex><pret>$ ^helt<adv>$ ^alene<adv>$ ^sørfra<adv>$ ^til<pr>$ ^Sameland<np><top>$^..<sent><clb>$

<e><p><l>Tine<s n="np"/><s n="ant"/><s n="f"/></l><r>Tine<s n="np"/><s n="ant"/><s n="f"/></r></p><par n="__np"/></e>

#22<det><qnt>un<pl> år gammel #Tine<np><f> #Chris<np><m> #Mathisen<np> #han<prn><p3><m><sg><nom> flyttet helt alene sørfra til Sameland.

$ echo Tine|apertium -d . nob-sme-morph
^Tine/Tine<np><ant><f>$^./.<sent><clb>$

Personlege pronomen

postchunk: ^jeg<prn><pers><p1><mf><sg><nom>$
nob fst:  ^jeg/jeg<n><nt><sg><ind>/jeg<n><nt><pl><ind>/jeg<prn><pers><p1><mf><sg><nom>$^./.<sent><clb>$

echo '^jeg<prn><pers><p1><mf><sg><nom>$'|lt-proc -g sme-nob.autogen.bin
jeg

Modes

Kevin har eit script for å lage alle stega i debugginga, alt her er ikkje sjekka (inn).

TILTAK

hfst vs bin

echo "Mii manahit oahpaheaddjiid."|apertium -d. sme-nob-disam|cg-conv -a
"<Mii>"
        "mii" prn ind sg nom
        "mii" prn itg sg nom
        "mii" prn rel sg nom
        "mun" prn pers p1 pl nom
"<manahit>"
        "manahit" vblex tv indic pres p1 pl
        "manahit" vblex tv indic pres p3 pl
        "manahit" vblex tv indic pret p2 sg
        "manahit" vblex tv inf
        "mannat" vblex iv der_h vblex tv indic pres p1 pl
        "mannat" vblex iv der_h vblex tv indic pres p3 pl
        "mannat" vblex iv der_h vblex tv indic pret p2 sg
        "mannat" vblex iv der_h vblex tv inf
"<oahpaheaddjiid>"
        "oahpaheaddji" n nomag sem_hum pl acc
        "oahpaheaddji" n nomag sem_hum pl gen
"<..>"
        ".." sent

Trond: $ echo "Mii manahit oahpaheaddjiid."|apertium -d. sme-nob-pretransfer
^Mii<prn><ind><sg><nom>$ ^manahit<vblex><tv><indic><pres><p1><pl>$ ^oahpaheaddji<n><nomag><sem_hum><pl><acc>$^..<sent>$
 echo "Mii manahit oahpaheaddjiid."|apertium -d. sme-nob
Hva mister lærere.

Lene:
echo "Mii manahit oahpaheaddjiid."|apertium -d. sme-nob-pretransfer
^Mun<prn><pers><p1><pl><nom><@SUBJ→>$ ^manahit<vblex><tv><indic><pres><p1><pl><@+FMAINV>$ ^oahpaheaddji<n><nomag><sem_hum><pl><acc><@←OBJ>$^..<sent>$
Vi mister lærere.

Kevin: $ echo "Mii manahit oahpaheaddjiid."|apertium -d. sme-nob-pretransfer
^Mun<prn><pers><p1><pl><nom><@SUBJ→>$ ^manahit<vblex><tv><indic><pres><p1><pl><@+FMAINV>$ ^oahpaheaddji<n><nomag><sem_hum><pl><acc><@←OBJ>$^..<sent>$

echo "Mii manahit oahpaheaddjiid."|apertium -d. sme-nob-disam|cg-conv -a
"<Mii>"
        "mun" prn pers p1 pl nom SELECT:4019:miiPersRight
        "¬mii" prn ind sg nom SELECT:4019:miiPersRight
        "¬mii" prn itg sg nom SELECT:4019:miiPersRight
        "¬mii" prn rel sg nom SELECT:4019:miiPersRight
"<manahit>"
        "manahit" vblex tv indic pres p1 pl SELECT:4695:VPl1IfMiiLeft MAP:7960:+FMAINV @+FMAINV
        "¬manahit" vblex tv indic pres p3 pl SELECT:4695:VPl1IfMiiLeft
        "¬manahit" vblex tv indic pret p2 sg SELECT:4695:VPl1IfMiiLeft
        "¬manahit" vblex tv inf SELECT:4695:VPl1IfMiiLeft
"<oahpaheaddjiid>"
        "oahpaheaddji" n nomag sem_hum pl acc SELECT:9422:AccTV2
        "¬oahpaheaddji" n nomag sem_hum pl gen SELECT:9422:AccTV2
"<..>"
        ".." sent

Kevin får:
 echo "Mii manahit oahpaheaddjiid."|apertium -d. sme-nob-disam|cg-conv -a
"<Mii>"
        "mun" prn pers p1 pl nom SELECT:4019:miiPersRight
        "¬mii" prn ind sg nom SELECT:4019:miiPersRight
        "¬mii" prn itg sg nom SELECT:4019:miiPersRight
        "¬mii" prn rel sg nom SELECT:4019:miiPersRight
"<manahit>"
        "manahit" vblex tv indic pres p1 pl SELECT:4695:VPl1IfMiiLeft MAP:7960:+FMAINV @+FMAINV
        "¬manahit" vblex tv indic pres p3 pl SELECT:4695:VPl1IfMiiLeft
        "¬manahit" vblex tv indic pret p2 sg SELECT:4695:VPl1IfMiiLeft
        "¬manahit" vblex tv inf SELECT:4695:VPl1IfMiiLeft
"<oahpaheaddjiid>"
        "oahpaheaddji" n nomag sem_hum pl acc SELECT:9422:AccTV2
        "¬oahpaheaddji" n nomag sem_hum pl gen SELECT:9422:AccTV2
"<..>"
        ".." sent

tf4-hsl-m0024:apertium-sme-nob trond$ echo "Mii manahit oahpaheaddjiid."|apertium -d ../../nursery/apertium-sme-smn  sme-smn-disam|cg-conv -a
"<Mii>"
        "mun" prn pers p1 pl nom SELECT:4004:miiPersRight
* **"mii" prn ind sg nom SELECT:4004**: miiPersRight
* **"mii" prn itg sg nom SELECT:4004**: miiPersRight
* **"mii" prn rel sg nom SELECT:4004**: miiPersRight
"<manahit>"
        "manahit" vblex tv indic pres p1 pl @+FMAINV SELECT:4680:VPl1IfMiiLeft MAP:7945:+FMAINV
* **"mannat" ex_vblex ex_iv der_h vblex tv indic pres p1 pl REMOVE:2268**: derV
* **"mannat" ex_vblex ex_iv der_h vblex tv indic pres p3 pl REMOVE:2268**: derV
* **"mannat" ex_vblex ex_iv der_h vblex tv indic pret p2 sg REMOVE:2268**: derV
* **"mannat" ex_vblex ex_iv der_h vblex tv inf REMOVE:2268**: derV
* **"manahit" vblex tv indic pres p3 pl SELECT:4680**: VPl1IfMiiLeft
* **"manahit" vblex tv indic pret p2 sg SELECT:4680**: VPl1IfMiiLeft
* **"manahit" vblex tv inf SELECT:4680**: VPl1IfMiiLeft
"<oahpaheaddjiid>"
        "oahpaheaddji" n nomag pl acc SELECT:9407:AccTV2
        "oahpaheaddji" n nomag sem_hum pl acc SELECT:9407:AccTV2
* **"oahpaheaddji" n nomag pl gen SELECT:9407**: AccTV2
* **"oahpaheaddji" n nomag sem_hum pl gen SELECT:9407**: AccTV2
"<.>"
        "." sent

"<.>"
        "." sent

Feilmeldingar

Kva input er det som gir dette? make Linjene 2131 og 2308 skriv ut utan pos-attributt.

Feilmeldinga mi kjem ikkje att etter kompilering, no får eg denne:

tf4-hsl-m0024:apertium-sme-nob trond$ touch apertium-sme-nob.sme-nob.t1x
tf4-hsl-m0024:apertium-sme-nob trond$ make
apertium-validate-transfer apertium-sme-nob.sme-nob.t1x
apertium-preprocess-transfer apertium-sme-nob.sme-nob.t1x sme-nob.t1x.bin
Warning (3625): Paths to rule 27 blocked by rule 21.

men rule 27 seier:

        <rule comment="REGLA: NOM (tl: NOM, VERB)
                   This is a catch-all rule; it's OK that some paths are blocked by previous rules.">

Regresjonstestar

Bruk

$ t/update-latest
$ svn diff t

for å køyra regresjonstestane og lagra resultatet i filer som er lagra i SVN – så får du ein diff med sist gong nokon køyrte testen.