Scandinavian MT project
- https://meta.wikimedia.org/wiki/Skanwiki/Tinget / https://meta.wikimedia.org/wiki/Skanwiki/Skanwikiprojekt_MT
- http://blog.wikimedia.org/2016/06/01/scandinavian-wikipedias-content-translation/ / http://blogg.wikimedia.no/2015/12/14/verktoy-for-tverrskandinavisk-maskinomsetjing/
Status
Note: UDHR was probably used during development on on nno/nob stuff.
Direction | Pair | Bidix | Coverage UDHR – Wiki |
Testvoc | Release date | Released | WER (-u) |
---|---|---|---|---|---|---|---|
nob-nno | nno-nob | 69,397 | 99.2% – 92.6% | ✓ | 10.71 %[1] | ||
nno-nob | nno-nob | 69,397 | 98.9% – 90.6% | ✓ | |||
dan-nob | dan-nor | 53,746 | 95.9% – 88.1% | ✓ | 1 februar 2016 | ✓ | 10.87 %[2] |
dan-nno | dan-nor | 53,746 | 92.7% – 87.3% | ✓ | 1 januar 2016 | ✓ | 13.64 %, 22.64 %[3] |
nob-dan | dan-nor | 53,746 | 98.6% – 91.7% | ✓ | ✓ | ||
nno-dan | dan-nor | 53,746 | 97.4% – 89.8% | ✓ | 1 april 2016 | ✓ | |
swe-nob | swe-nor | 8,920 | 94.8% – 87.3% | 17 mai 2016 | |||
swe-nno | swe-nor | 8,920 | 93.8% – 87.0% | 17 mai 2016 | |||
nob-swe | swe-nor | 8,920 | 97.1% – 89.7% | 17 mai 2016 | |||
nno-swe | swe-nor | 8,920 | 96.6% – 87.4% | 17 mai 2016 | |||
swe-dan | swe-dan | 17,551 | 88.0% – 83.7% | 1 mars 2016 | ✓ | 31 %[4] | |
dan-swe | swe-dan | 17,551 | 90.4% – 82.9% | 1 mars 2016 | ✓ |
Stats from stemcounterbot
(bidix stems buggy for swe-pairs)
Pair | Bidix | t1x rules | lrx rules |
---|---|---|---|
nno-nob | 69,479 | nno→nob: 22, nob→nno: 85 | nno→nob: , nob→nno: |
dan-nor | 54,742 | dan→nno: 31, dan→nob: 24, nor→dan: 22 | dan→nno: , dan→nob: , nor→dan: |
swe-nor | 66,856 | swe→nno: 30, swe→nob: 26, nor→swe: 27 | swe→nno: , swe→nob: , nor→swe: |
swe-dan | 21,791 | swe→dan: 24, dan→swe: 26 | swe→dan: , dan→swe: |
Evaluation
- Swedish: kalmar.txt (1119 tokens)
- Danish: venedig.txt (1363 tokens)
- Norwegian Nynorsk:
- Norwegian Bokmål: