The Divvun and Giellatekno teams build language technology aimed at minority and indigenous languages
View GiellaLT on GitHub divvungiellatekno/giellalt.uit.no
Til stades: Lene, Linda, Sara (i starten), Thomas, Trond, Sjur
Ord med tvetydig eller problematisk fokuspartikkelanalyse:
"<dálka>"
"dálka" N Sem/Dummytag Sg Nom <W:0.0> @HNOUN MAP:22636:hnounNom
* **"dálkat" V IV PrsPrc <W:0.0> REMOVE:3963**: PrsPrc
"<dat>"
"dat" Pcle <W:0.0> @PCLE MAP:21642:r16
Ny analyse, berre hfst-tokenise:
echo "Son čohkkáhan dáppe." | hfst-tokenise -g tokeniser-gramcheck-gt-desc.pmhfst
"<Son>"
"son" Pcle <W:0.0>
"son" Pron Pers Sg3 Nom <W:0.0>
:
"<čohkkáhan>"
"čohkkáhit" V TV Actio Gen <W:0.0>
"čohkkáhit" V TV Actio Nom <W:0.0>
"čohkkáhit" V TV Ind Prt ConNeg <W:0.0>
"čohkkáhit" V TV PrfPrc <W:0.0>
"čohkkát" Ex/V IV Der/h V Actio Gen <W:0.0>
"čohkkát" Ex/V IV Der/h V Actio Nom <W:0.0>
"čohkkát" Ex/V IV Der/h V Ind Prs Sg1 <W:0.0>
"čohkkát" Ex/V IV Der/h V Ind Prt ConNeg <W:0.0>
"čohkkát" Ex/V IV Der/h V PrfPrc <W:0.0>
"#" Foc/han <W:0.0> "<han>"
"čohkkát" V IV Imprt Sg2 <W:0.0> "<čohkká>"
"#" Foc/han <W:0.0> "<han>"
"čohkkát" V IV Ind Prs ConNeg <W:0.0> "<čohkká>"
"#" Foc/han <W:0.0> "<han>"
"čohkkát" V IV Ind Prs Sg3 <W:0.0> "<čohkká>"
"#" Foc/han <W:0.0> "<han>"
"čohkkát" V IV VGen <W:0.0> "<čohkká>"
:
"<dáppe>"
"dáppe" Adv Sem/Plc <W:0.0>
"<.>"
"." CLB <W:0.0>
:\n
Ny analyse etter cg-mwesplit, klitisert:
$ echo "Son čohkkáhan dáppe." | hfst-tokenise -g tokeniser-gramcheck-gt-desc.pmhfst | vislcg3 -t -g mwe-dis.cg3 | cg-mwesplit
"<Son>"
"son" Pcle <W:0.0>
"son" Pron Pers Sg3 Nom <W:0.0>
:
"<čohkká>"
"čohkkát" V IV Ind Prs Sg3 <W:0.0> SELECT:2146
"<han>"
"#" Foc/han <W:0.0>
* ** "čohkkáhit" V TV Actio Gen <W:0.0> SELECT**: 2146
* ** "čohkkáhit" V TV Actio Nom <W:0.0> SELECT**: 2146
* ** "čohkkáhit" V TV Ind Prt ConNeg <W:0.0> SELECT**: 2146
* ** "čohkkáhit" V TV PrfPrc <W:0.0> SELECT**: 2146
* ** "čohkkát" Ex/V IV Der/h V Actio Gen <W:0.0> SELECT**: 2146
* ** "čohkkát" Ex/V IV Der/h V Actio Nom <W:0.0> SELECT**: 2146
* ** "čohkkát" Ex/V IV Der/h V Ind Prs Sg1 <W:0.0> SELECT**: 2146
* ** "čohkkát" Ex/V IV Der/h V Ind Prt ConNeg <W:0.0> SELECT**: 2146
* ** "čohkkát" Ex/V IV Der/h V PrfPrc <W:0.0> SELECT**: 2146
* ** "#" Foc/han <W**: 0.0> "<han>"
* ** "čohkkát" V IV Imprt Sg2 <W:0.0> "<čohkká>" SELECT**: 2146
* ** "#" Foc/han <W**: 0.0> "<han>"
* ** "čohkkát" V IV Ind Prs ConNeg <W:0.0> "<čohkká>" SELECT**: 2146
* ** "#" Foc/han <W**: 0.0> "<han>"
* ** "čohkkát" V IV VGen <W:0.0> "<čohkká>" SELECT**: 2146
:
"<dáppe>"
"dáppe" Adv Sem/Plc <W:0.0>
"<.>"
"." CLB <W:0.0>
:\n
Ny analyse etter cg-mwesplit, ikkje klitisert:
$ echo "Son čohkká han dáppe." | hfst-tokenise -g tokeniser-gramcheck-gt-desc.pmhfst | vislcg3 -t -g mwe-dis.cg3 | cg-mwesplit
"<Son>"
"son" Pcle <W:0.0>
"son" Pron Pers Sg3 Nom <W:0.0>
:
"<čohkká>"
"čohkkát" V IV Ind Prs Sg3 <W:0.0> SELECT:2146
"< han>"
" " Foc/han <W:0.0>
* ** " " Foc/han <W**: 0.0> "< han>"
* ** "čohkkát" V IV Imprt Sg2 <W:0.0> "<čohkká>" SELECT**: 2146
* ** " " Foc/han <W**: 0.0> "< han>"
* ** "čohkkát" V IV Ind Prs ConNeg <W:0.0> "<čohkká>" SELECT**: 2146
* ** " " Foc/han <W**: 0.0> "< han>"
* ** "čohkkát" V IV VGen <W:0.0> "<čohkká>" SELECT**: 2146
:
"<dáppe>"
"dáppe" Adv Sem/Plc <W:0.0>
"<.>"
"." CLB <W:0.0>
:\n
Gamal analyse:
$ echo "Son čohkká han dáppe." | preprocess | hfst-lookup -q ../../src/analyser-disamb-gt-desc.hfst| cut -f1-2 | lookup2cg
"<Son>"
"son" Pcle
"son" Pron Pers Sg3 Nom
"<čohkká>"
"čohkkát" V IV Ind Prs Sg3
"čohkkát" V IV VGen
"čohkkát" V IV Ind Prs ConNeg
"čohkkát" V IV Imprt Sg2
"čohkkát" V IV Imprt ConNeg
"<han>"
"han" Pcle
"<dáppe>"
"dáppe" Adv Sem/Plc
"<.>"
"." CLB
Spørsmål:
Svar:
Jf. analysen av desse to — láibebiila og čohkkáhan:
usme:
láibebiila láibi+N+Cmp/SgNom+Cmp#biila+N+Sg+Nom
echo "láibebiila" | hfst-tokenise -g tokeniser-gramcheck-gt-desc.pmhfst
"<láibebiila>"
"biila" N Sem/Veh Sg Nom <W:10.0>
"láibi" N Sem/Food Cmp/SgNom Cmp <W:10.0>
usme:
čohkkáhan čohkkát+V+IV+Ind+Prs+Sg3+Foc/han
hfst-tokenise:
"han" Pcle <W:0.0> "<han>"
"čohkkát" V IV Imprt Sg2 <W:0.0> "<čohkká>"
For usme-analysen kunne vi i staden ha:
čohkkát+V+IV+Ind+Prs+Sg3#han+Pcle+Foc
Her ser vi korleis biila og han blir lemma i hfst-tokenise, og indirekte korleis det kunne ha vorte det for čohkkahan.
Nickel/Sammallahti s. 205: «Modalpartikler»
Partikler:
---------------------------------------
Pcle:
2. posisjon eller fokus/polaritet: Pcle
leago +Pcle+Qst <cl> evt: +Pcle+Qst+Clt
lea go +Pcle+Qst
leage +Pcle+Pol <cl>
lea ge +Pcle+Pol
leago +Pcle+Qst <cl>
lea go +Pcle+Qst
leaba +Pcle+Foc <cl>
lea ba +Pcle+Foc
-------------------------------------
Ikkje 2. posisjon eller fokus: Adverb
aitto Pcle <== Adv Cls Sem/xxx?
odne Adv
manne+Adv+Subqst
dáppe+Adv+Sem/Plc
dalle+Adv+Sem/Time
čččč+Adv+Sem/Plc_Time ?
-------------------------------------
husmedis
mihkkege
mihkkege mihkke+Pron+Indef+Sg+Nom+Foc/Neg-ge 0,000000
mihkkege mihkke+Pron+Indef+Sg+Nom+Foc/Pos-ge 0,000000
Pron Pers
N NomAg
Legge inn LIA-etntrŧyer i fil/filer?
grep LIA src/morphology/stems/* |wc -l
2175
Begrunnelse: Dette gir mulighet for å kompilere en egen HFST for LIA-analyse, og vi risikerer ikke at ordene gir for mange analyser i andre sammenhenger
REMOVE:longest-match SUB:1 ("<.*>"r)(NEGATE 0 Err/Lex);
SELECT V + Sg3 (0 Foc/han LINK -1 Pers + Sg3 + Nom) ;
Dette er måten å referere til underlesingar på:
# SUBREADINGS:
# ------------
bil 0, -3
kake 1, -2
oste 2, -1
# oste#kake#bil
# 0 = bil
# 0/0 = bil
# 0/1 = kake
# 0/2 = oste
# 0/-1 = oste
# 0/-2 = kake
# 0/-3 = bil
1. Til venstre for "/": posisjon i setninga
1. Til högre for "/": posisjon i stakken, rekna ovafrå (positiv) eller nedanfrå (negativ