160105

• Bidix improvement
• Progress
• Next meeting

Bidix improvement

We will improve bidix by:

• frequency sorting: Ciprian
• checking sme lemmas: Lene
• checking smj lemmas:
  • in dev: sh bidix-sanity.sh > sanityoutput (Sandra)
  • if same structure in sme and smj, one can remove the word pair instead of lexicalizing it
• adding word pairs from Lenes giza-experiments in 2008: Lene makes a csv-file, Sandra proofread
• checking sme verbs IV vs TV: done
• change back to Po, Pr and Adv in smj-lexc
• add transfer word pairs from termlists (sme-nob + nob-smj) and compare them to bidix:
  • Give priority to sme-nob/nob-smj (where nob-smj = Kintel) pairs that are already in bidix, and lift they over in a priority bidix list
  • evaluate sme-smj-bidix not in sme-nob/nob-smj and sme-nob/nob-smj not in sme-smj-bidix separately, after the seeded gang
  • For the sme-smj-bidix residue: Give priority to sme-smj pairs where smj are already in FST
• make missing list from sme-corpus: Trond

Progress

Pronouns

There are two files with sentences with pronouns in texts/ for comparing and improving word pairs in bidix and harmonising tags in FST

Transfer rules

We will utilize on work done with other sme-smX pairs by

• using the same names on variabels and sets
• document relevant rules in a document on smX MT web-page
• suggest for Francis to hold regular meetings for discussing transfer rules, and improvements

Next meeting

Friday 8.1 kl 13 (Sandra and Lene)

We'll have meeting once a week (but next meeting will be the last week of January)