Language Technology at UiT

The Divvun and Giellatekno teams build language technology aimed at minority and indigenous languages

View GiellaLT on GitHub divvungiellatekno/giellalt.uit.no

Meeting setup

Agenda

Cf. one of the following, depending on context:

Opening, agenda review, participants

Opened at 09:48.

Present: Børre, Ciprian, David, Sjur, Thomas, Tomi

Absent: Jovsset, Trond

Agenda accepted as is.

Updated task status since last meeting

Børre

Ciprian

David

Jovsset

Lene

Per-Eric

Saara

Sjur

Thomas

Tomi

Trond

Pedagogical software online

Meeting memos can be found at [http://giellatekno.uit.no/ped/index.html#Meeting+memos]

TODO:

Corpus gathering

David added one book, originally in Swedish, translated into sma as well as sme and smj. Added sma text, will add original, and will ask for permission to add thnowe other two translations as well.

Contract feedback: the contract is too long, and too heavy. We need a short, or “light”, version of the contract.

Processing steps for Sámi-Norwegian place name pairs from SK:

TODO:

Promoting Divvun

Thomas and Tomi were at Rica Hotell Tromsø last week, presenting the Divvun project for the SGL (Norwegian normative group, all but sma was represented). They were very positively surprised, and seemed interested. They even wanted to send texts to us.

Learning experience for us: even informed people do not really get what it is all about until we meet them. Which means that we need to meet more people, to get more people to use our tools and to send us texts.

Thomas talked about the normative work, especially the smj normative work. It is very unclear how loanwords should be normalised. Mentioned that he had made some decisions himself (with clear examples), and got positive response on that.

TODO:

Future plans, directions and ideas

See a separate document in plan/strat/5year.jspwiki.

Infrastructure

To accomodate future enhancements in different directions (in rough order of importance):

  1. test bench for all parts of our language technology efforts
    1. test bench enhanced, but not yet complete
  2. more modularised make / build infra (prepare for smn, sms, sjd, others)
  3. merge gt, kt and st into one
  4. set up the Leopard Server features for collaborative support:
    1. permanent chat rooms
    2. stored (and indexed) chat transcripts of the chat rooms
    3. iCal server / group calendars
    4. wiki
  5. wiki? on G5 (is part of Leopard Server) or other web-based documentation
  6. improve Forrest stability and i18n support ( the divvun crashes)
    1. Sjur has been working on better i18n and pdf rendering
    2. Børre has some ideas for getting back to serving static html files
  7. reorganise the documentation:
    1. differ between target groups
    2. get better grouping
    3. decide what to write in forrest and what in wiki (cf. Apertium and [http://xixona.dlsi.ua.es/apertium/]) for a similar split)
    4. update/add missing parts
  8. migrate lexc lexicons to XML, splitting the task
    1. Name lexica (the Name project)
    2. Dictionaries (already in XML, task is to integrate them)
    3. At least migrate the lexc open POSes (Komi as a pilot case)
  9. change the look of the documentation web
  10. sfst?
    1. replacement for xfst -> see [omorf | http://www.ling.helsinki.fi/kieliteknologia/tutkimus/omor/index.shtml], omorf wiki and omorfi
    2. replacement for hunspell/open-source proofing tools
  11. investigate the NSIS installer, potentially replacing the InstallShield package from Polderland
  12. corpus content moved to Max Planck repositories? Norsk språkbank?
  13. update infrastructure to allow content-restricted spellers for special target groups

SVN issues:

TODO:

Linguistics

North Sámi

(nothing new, see proofing bugs below)

http://giellatekno.uit.no/bugzilla/show_bug.cgi?id=711

Lule Sámi

(nothing new, see proofing bugs below)

TODO:

South Sámi

TODO:

Name lexicon/risten.no infrastructure

TODO:

  1. fix i18n bug in risten.no/G5 (so they will work without the proper locale request) (Sjur)
  2. fix bugs in lexc2xml; add comments to the log element (Saara)
  3. finish first version of the editing (Sjur)
  4. test editing of the xml files. If ok, then: (Sjur, Thomas, Trond)
  5. make terms-smX.xml <=== automatically from propernoun-sme-lex.xml (add nob as well) (the morphological section should be kept intact, in e.g. propernoun-sme-morph.txt) (Sjur, Saara)
  6. convert propernoun-($lang)-lex.txt to a derived file from common xml files (Sjur, Tomi, Saara)
  7. implement data synchronisation between risten.no and the cvs repo, and possibly other servers (ie the G5 as an alternative server to the public risten.no - it might be faster and better suited than the official one; also local installations could be treated the same way)
  8. start to use the xml file as source file
  9. clean terms-sme.xml such that all names have the correct tag for their use (e.g. @type=secondary) (Thomas, linguists)
  10. merge placenames which are errouneously in different entries: e.g. Helsinki, Helsingfors, Helsset (linguists)
  11. publish the name lexicon on risten.no (Sjur)
  12. add missing parallel names for placenames (linguists)
  13. add informative links between first names like Niillas and Nils (linguists)

Dictionaries

nob-sme dictionary works, ie compilation and building works as it should. Ciprian has also solved the paradigm generation issue on his computer, which means morphological intelligence should be back in sme-nob.

To get the paradigm generation working, clean everything and make a new automata and it should work.

September:

TODO:

Proofing tools

Hunspell

In our process to make hunspell tools, makealias, a script that compacts the hunspell dictionaries, has been incorporated (makealias is a shell script available as part of the source distribution). The resulting files seem to crash OO.org 3.0rcX at least on Windows and OS X.

The same files work well on OO.org 2.4 on Linux and Windows, and also as systemwide spellcheckers for both Gnome and KDE4 applications. Using the non-compacted hunspell dictionaries on OS X and OO.org 3.0 works well. On Windows (in a virtual machine and 256MB RAM) and OO.org 3.0, the spelling process stalls, probably due to too little RAM. Børre will test it again with more RAM.

TODO:

OpenXSpell

Working everywhere (almost) in MacOS X applications, provides Sámi spell checking system-wide through a hunspell back-end. The main problem is that it isn’t yet UTF-8 safe. OpenXSPell

TODO:

Testing

Spelling Error Markup

TODO:

Speller testing

We need a test file for testing the speller behaviour on a defined set of word constructions: compounds of all types, derivations, inflections, and combinations of those.

TODO:

Speller updates

TODO:

Speller bugs

List of bugs returned from Polderland: 621, 630, 652, 656, 676.

Open issues based on test results:

sme

Version: Davvisámi, version 1.0.1, 2008-09-25

smj

Version: Julevsáme, version 1.0.1, 2008-09-15

TODO:

Hyphenator bugs

Open issues based on test results :

sme

Lexicon version: Davvisámi, version 1.0.1, 2008-04-01

smj

Lexicon version: Julevsáme, version 1.0.1, 2008-04-01

TODO:

InDesign tools

Nothing new.

Releases

TODO:

Other

Forthcoming Sámi allaskuvla conference

Info on the conference site.

Title submitted is: Goallosteapmi ja suorggideapmi Divvun-reaidduin

Registration: [http://www.viaregi.no/registration/deltagerweb.aspx?kid=3440&aid=8700]

Registration deadline is September 10.

TODO:

Presentation at SGL, Rica Hotell Tromsø

Done, see report further up. No CD’s showed up, though :(

Text to speech

We are planning a North Sámi text to speech. Work on the text-to-transcription (ttt) component has begun, and for the transcription-to-sound we cooperate with U Hki. For a forthcoming october demo we have two alternative tracks:

  1. Do the ttt with quasi-Finnish orthography as output, and plug it onto the Finnish ttt (the Sámi-as-a-Finnish-priest solution) (pro: the web demo is there, so the demo is within reach; con: several phonemes are missing)
  2. Do the ttt with SAMPA as output, and read in 200 sentences, and have Hki generate a voice based upon that (pro: probably better result, head start on the project; con: we risk that is not be done by
  3. Do both in parallel. (pro: safety first, not much double work; con: well, the double work in the phon-sme.xfst file)

The best thing would probably be to do both tracks, to be on the safe side.

echo "23847 ja de mun ipmirdán ja 12° ja §12 ja 23,2 ja 23,- ja 23-23" | jietna.sh
kuokː.hte.lo.gi kolə.bmɑ tuː.hɑːh kɑːvht.tsiː tʃuo.ðiː ɲeæl.lje lo.giː tʃie.dʃɑ
jɑ te mun ip.mir.dɑːn jɑ kuokː.hte nup.pe.lohː.kɑːj grɑː.dɑ jɑ pɑ.rɑː.grɑː.fɑ...

TODO:

Corpus contracts + open source

Postponed until the svn repository is fully functional (it is too open now).

Next meeting, closing

The next meeting is 13.10.2008, 9.30 Norwegian time.

The meeting was closed at 11:51.

Appendix - task lists for the next five days

Boerre

Ciprian

David

Jovsset

Lene

Per-Eric

Saara

Sjur

Thomas

Tomi

Trond