Pending tasks
- Tag a corpus with IceTagger and train the
apertium-tagger
- Try and convert some IceTagger constraint rules to work in constraint grammar
- Post-edit automatically-generated bilingual dictionaries
- Use IceParser to parse a corpus and extract the most frequent patterns in terms of chunks/phrases (lists of coarse POS tags) and phrase patterns (in terms of chunks/phrases).
- Merge analysed corpus (IceMorphy full-form list) with Apertium dictionary — will require matching partial information to paradigms... perhaps use extract ?
Resources
Corpora
- Mediawiki l10n, KDE4, OpenSubtitles, etc. — from OPUS (~60k sentences)
- Parliamentary Acts (?? sentences)
Bilingual dictionaries
Example phrase
- Hver maður er borinn frjáls og jafn öðrum að virðingu og réttindum.
IceFormat
Hver foken maður nken er sfg3en
borinn sþgken frjáls lkensf
og c jafn aa öðrum fokfþ að c
virðingu nveþ og c réttindum nhfþ . .
{*SUBJ> [NP Hver foken maður nken NP] *SUBJ>}
[VPb er sfg3en VPb]
{*COMP< [VPp borinn sþgken VPp] *COMP<}
{*COMP< [APs [AP frjáls lkensf AP] [CP og c CP] [AP jafn lkensf AP] APs] *COMP<}
[NP öðrum fokfþ NP]
[SCP að c SCP]
[NPs [NP virðingu nveþ NP] [CP og c CP] [NP réttindum nhfþ NP] NPs]
Apertium
^Hver<prn><ind><m><sg><nom>$ ^maður<n><m><sg><nom><ind>$ ^vera<vbser><pri><p3><sg>$
^bera<vblex><pp><m><sg><nom>$ ^frjáls<adj><sta><pst><m><sg><nom>$
^og<cnjcoo>$ ^jafn<adj><sta><pst><m><sg><nom>$
^annar<prn><ind><m><pl><dat>$ ^að<pr>$
^virðing<n><f><sg><dat><def>$ ^og<cnjcoo>$ ^réttindi<n><nt><pl><dat><ind>$ ^.<sent>$
^prn_nom<SN><@SUBJ→>{^Hver<prn><ind><m><sg><nom>$ ^maður<n><m><sg><nom><ind>$}$
^verb<SV>{^vera<vbser><pri><p3><sg>$ ^bera<vblex><pp><m><sg><nom>$}$
See also
