GiellaLT provides an infrastructure for rule-based language technology aimed at minority and indigenous languages, and streamlines building anything from keyboards to speech technology.

View GiellaLT on GitHub

Page Content

OmegaT and machine translation

OmegaT as an MT platform

OmegaT is one of many computer-assisted translation (CAT) programs, i.e. programs that help you translate documents from one language to another. OmegaT is open source and it is thus possible to adapt it to translation to and from Saami languages. Most of [our resources for computer-assisted translation](../../tm/TranslationMemory.html] can be used by all CAT programs, though (the exception is machine translation, that works only for OmegaT).


The user documentation page for OmegaT refers to installation and user documentation, and can be found here:

Internal plans for OmegaT development

What follows are our thoughts for developing CAT for Saami.


The idea is to offer a set of ready-made folders, perhaps in two different formats:

  1. as a one-time downloading of a zipped file archive
  2. as svn checkout (via Tortoise on Windows) for access to updates

For the time being, the folders are at [].

The idea is to put the following resources into the following subdirectories:

The source and target folders are given svn ignore status, as we develop the folders we should determine what other files to ignore and what to share.

The language pairs

The language pairs are of three types:

  1. smesmn, smesmj, smesmn: The main thing here is MT, glossaries and dictionaries are less interesting since they are already in bidix, and since we do not have a OmegaT-compatible tokenizer to look up inflected words.
  2. nobsme, nobsmj, nobsma, finsme, finsmn, finsms: Here we have no MT (except for finsme, which is not much developed). The focus here is on glossaries (fad project, etc.) and translation memory
  3. smasme, smjsme, smnsme, smenob: these we ignore in OmegaT for now. They are mainly made for understanding, not for text production.


  1. Add glossaries (done for nobsmX)
  2. Improve mt (done for nobsmX)
  3. Develop segmentation.conf (in progress for sme)
  4. Test and evaluate

Future plans

Adding more resources:

HFST Tokenizer

You can get the hfst tokenizer ready compiled. You need to download:

And put them into ~/Library/Preferences/OmegaT/plugins (create the dir if it’s not there). Tested and works with OmegaT 4.x.

Hfst tokenizer source is at github