151216

MT-čoahkkin 16.12.2015

Ciprian, Erika, Lene, Maja, Marja-Liisa, Trond

Áššelistu

  1. geahččat dokumentašuvnna (the list) (Maja sáhttá leat mielde dán oasis)
    1. mo gánnáha bargat geavatlaččat máŋggaid lásiiguin
  2. Miina ja Neeta barggut
  3. Prošeavtta progrešuvdna guovvamánu rádjai

Geahččat dokumentašuvnna (the list)

Dokumentašuvnnas leat barggut čilgejuvvon dárkilit. Lea dehálaš buoridit dan jus mihkkege lea eahpečielggas.

Apertium-gohččumat:

Ovdamearka: MT vuogádat geahččala suorggidahttit pyeri adjektiivva, mii oažžu #-mearkka. Dát mearkkaša ahte FST ii dovdda dan sojaheami dahje ii dovdda sáni ollenge.
Vuosttaš jorgalus lea dahkkon cat texts/vajaldumjuovlah.sme.txt |apertium -d . sme-smn -gohččumiin, nubbi lea dahkkon cat texts/vajaldumjuovlah.sme.txt |apertium -d . sme-smn-dgen -gohččumiin:

  • piärgukáhhoid #pyeri ko vuoššum kyelikáhhoid
  • piärgukáhhoid #pyeri<adj><comp><der_avt><adv> ko vuoššum kyelikáhhoid

Dokumentašuvnnas leat čilgejuvvon njeallje bidix-bargovuogi

Missinglistui leat čohkkejuvvon buot sániid mat ožžot nástti go mii jorgalit buot teavsttaid mat leat texts-máhpas. Násti mearkkaša ahte sátni ii leat bidix-fiillas, dahje ahte das lea boasttu sátneluohkká dahje (jus vearba) IV/TV bidixas.

Mii sáhttit skriptet buot propernouns mat leat sme-propernouns-fiillas bidixii, muhto mii fertet maiddái mearridit mo dat galget sojahuvvot (dat mat eai galggaše jorgaluvvot anárašgillii).

Miina ja Neeta barggut

  • Add words to bidix
  • Check word pairs in bidix

Trond to discuss + Lene to look at tasks for them.

Prošeavtta progrešuvdna guovvamánu rádjai

  • Article for Szeged (revision), Final Version Due Dec 20, 2015
    • art/2016/iwclul/smnpaper/smnfst.pdf (edit in smnfst.tex)
    • art/2016/iwclul/smnpaper/review.txt (comment here: %% is the review, TT etc is us)
  • Talk in Szeged, the conference is Jan 20, 2016 - Jan 21, 2016
  • Talk on MT
    • evaluation: translate a text and ask some people to correct them, and then calculate WER (word error rate)
    • The goal before the evaluation:
      • work with missinglist
      • improve pronouns (send sentences to IM for tranlation, this week)
      • get rid of most #: s (e.g. correct PoS, tags, add to FST)
    • deadline for improving MT-system before the evaluation 19.1 ?
  • Talk on 100w
  • Talk on FST (formal workshop)
  • Week in Tromsø in February -- including the weekend before the conference

Next meetings:

  • Friday on the article
  • Monday 21st at 0900 on MT