Difference between revisions of "North Saami and Lule Saami"
Jump to navigation
Jump to search
(→TODO) |
|||
Line 19: | Line 19: | ||
* Re-train the HMM-based POS tagger on a Sami corpus. |
* Re-train the HMM-based POS tagger on a Sami corpus. |
||
* Closed categories in sme analyser |
* Closed categories in sme analyser |
||
==Reminders== |
|||
* In the transfer rule files, don't forget to escape the '+' character in tags, for example: '''no:''' <code><attr-item tags="@\+FMAINV"/></code> , '''yes:''' <code><attr-item tags="@\+FMAINV"/></code> |
|||
==Testing== |
==Testing== |
Revision as of 10:26, 7 October 2008
Contents |
Files
apertium-sme-smj.sme.dix
— Northern Sami transducerapertium-sme-smj.sme-smj.dix
— Transfer lexiconapertium-sme-smj.smj.dix
— Lule Sami transducerapertium-sme-smj.sme-smj.rlx
— Constraint grammarapertium-sme-smj.sme-smj.t1x
— Transfer rule file (level 1 -- Local re-ordering, chunking)apertium-sme-smj.sme-smj.t2x
— Transfer rule file (level 2 -- Phrase and chunk re-ordering)apertium-sme-smj.sme-smj.t3x
— Transfer rule file (level 3 -- Final touches)
TODO
- Mapped tags in the CG use special characters in Apertium, for example '>' (used for delimiting tags) and '-' (causes problems with pretransfer). These should be replaced somehow.
- Example:
^Wikipedia<N><Prop><Sg><Nom><@SUBJ>>$
or^prošeakta<N><Sg><Nom><@<SPRED>$
- This comes from the CG tag @SUBJ>
- Example:
- Re-train the HMM-based POS tagger on a Sami corpus.
- Closed categories in sme analyser
Reminders
- In the transfer rule files, don't forget to escape the '+' character in tags, for example: no:
<attr-item tags="@\+FMAINV"/>
, yes:<attr-item tags="@\+FMAINV"/>
Testing
- Analysing some Northern Sami text
$ echo "Wikipedia lea máŋggagielat prošeakta man ulbmilin lea ráhkadit almmolaš diehtosátnegirjji gosa \ gii beare sáhttá čállit artihkkaliid." | lt-proc sme-smj.automorf.bin ^Wikipedia/Wikipedia<N><Prop><Sg><Nom>/Wikipedia<N><Prop><Sg><Gen>/Wikipedia<N><Prop><Sg><Acc>$ ^lea/leat<V><IV><Ind><Prs><Sg3>$ ^máŋggagielat/*máŋggagielat$ ^prošeakta/prošeakta<N><Sg><Nom>$ ^man/man<ADV>$ ^ulbmilin/*ulbmilin$ ^lea/leat<V><IV><Ind><Prs><Sg3>$ ^ráhkadit/ráhkadit<V><TV><Inf>/ráhkadit<V><TV><Ind><Prs><Pl3>/ráhkadit<V><TV><Ind><Prt><Sg2>$ ^almmolaš/almmolaš<A><Attr>/almmolaš<A><Sg><Nom>$ ^diehtosátnegirjji/*diehtosátnegirjji$ ^gosa/gosa<ADV>/gossat<V><IV><VGen>/gossat<V><IV><Imprt><Prs><ConNeg>/gossat<V><IV><Imprt><Prs><Sg2>/gossat<V><IV><Ind><Prs><ConNeg>$ ^gii/*gii$ ^beare/beare<ADV>$ ^sáhttá/sáhttit<V><IV><Ind><Prs><Sg3>$ ^čállit/čállit<V><TV><Inf>/čállit<V><TV><Ind><Prs><Pl1>$ ^artihkkaliid/*artihkkaliid$.
- Disambiguating and annotating text with Constraint grammar
$ echo "Wikipedia lea máŋggagielat prošeakta man ulbmilin lea ráhkadit almmolaš diehtosátnegirjji gosa \ gii beare sáhttá čállit artihkkaliid." | lt-proc sme-smj.automorf.bin | cg-proc sme-smj.rlx.bin ^Wikipedia/Wikipedia<N><Prop><Sg><Nom><@SUBJ%>$ ^lea/leat<V><IV><Ind><Prs><Sg3><@+FMAINV>$ ^máŋggagielat/*máŋggagielat$ ^prošeakta/prošeakta<N><Sg><Nom><@<SPRED>$ ^man/man<ADV>$ ^ulbmilin/*ulbmilin$ ^lea/leat<V><IV><Ind><Prs><Sg3><@+FMAINV>$ ^ráhkadit/ráhkadit<V><TV><Inf><@-FMAINV>$ ^almmolaš/almmolaš<A><Sg><Nom><@%SUBJ>$ ^diehtosátnegirjji/*diehtosátnegirjji$ ^gosa/gosa<ADV>$ ^gii/*gii$ ^beare/beare<ADV>$ ^sáhttá/sáhttit<V><IV><Ind><Prs><Sg3><@+FAUXV>$ ^čállit/čállit<V><TV><Inf><@-FMAINV>$ ^artihkkaliid/*artihkkaliid$
- Finishing off the disambiguation with Apertium's HMM tagger
$ echo "Wikipedia lea máŋggagielat prošeakta man ulbmilin lea ráhkadit almmolaš diehtosátnegirjji gosa \ gii beare sáhttá čállit artihkkaliid." | lt-proc sme-smj.automorf.bin | cg-proc sme-smj.rlx.bin | apertium-tagger -g sme-smj.prob ^Wikipedia<N><Prop><Sg><Nom><@SUBJ%>$ ^leat<V><IV><Ind><Prs><Sg3><@+FMAINV>$ ^*máŋggagielat$ ^prošeakta<N><Sg><Nom><@<SPRED>$ ^man<ADV>$ ^*ulbmilin$ ^leat<V><IV><Ind><Prs><Sg3><@+FMAINV>$ ^ráhkadit<V><TV><Inf><@-FMAINV>$ ^almmolaš<A><Sg><Nom><@%SUBJ>$ ^*diehtosátnegirjji$ ^gosa<ADV>$ ^*gii$ ^beare<ADV>$ ^sáhttit<V><IV><Ind><Prs><Sg3><@+FAUXV>$ ^čállit<V><TV><Inf><@-FMAINV>$ ^*artihkkaliid$
- Applying lexical transfer and chunking
$ echo "Wikipedia lea máŋggagielat prošeakta man ulbmilin lea ráhkadit almmolaš diehtosátnegirjji gosa \ gii beare sáhttá čállit artihkkaliid." | lt-proc sme-smj.automorf.bin | cg-proc sme-smj.rlx.bin | apertium-tagger -g sme-smj.prob | \ apertium-transfer apertium-sme-smj.sme-smj.t1x sme-smj.t1x.bin sme-smj.autobil.bin ^nom<SN><@SUBJ%><Sg><Nom>{^Wikipedia<N><Prop><Sg><Nom>$}$ ^default{^liehket<V><IV><Ind><Prs><Sg3><@+FMAINV>$}$ ^unknown{^*máŋggagielat$}$ ^nom<SN><Sg><Nom>{^prosjækta<N><Sg><Nom>$}$ ^default{^man<ADV>$}$ ^default{^@ulbmil<N><Ess><@SPRED>>$}$ ^default{^liehket<V><IV><Ind><Prs><Sg3><@+FMAINV>$}$ ^default{^dahkat<V><TV><Inf><@-FMAINV>$}$ ^default{^almulasj<A><Sg><Nom><@%SUBJ>$}$ ^unknown{^*diehtosátnegirjji$}$ ^default{^<ADV>$}$ ^unknown{^*gii$}$ ^default{^@beare<ADV>$}$ ^default{^sáhttet<V><IV><Ind><Prs><Sg3><@+FAUXV>$}$ ^default{^tjállet<V><TV><Inf><@-FMAINV>$}$ ^unknown{^*artihkkaliid$}$.