Linda Wiechetek, Francis M. Tyers, and Thomas Omma: Shooting at Flies in the Dark: Rule-Based Lexical Selection for a Minority Language Pair

Adding a lexical selection module to a ruled based MT system for translation from North Sami to Lule Sami.

• All the resources and source codes are downloadable.
• Grammar rules of the minority languages (which can be dying out) are formalised and thus preserved.

Many real-world examples from discussed languages make the understanding of the paper easier. -MK-

• The approach needs a lot of human work.
• The evaluation is unclear and subjective. Sentence pairs which the authors considered unequivalent were removed from the data. There is no baseline approach to compare with.

1. Where does the semantic information come from? Is it from manual annotation?
2. Assimilation vs. dissemination. Does it make a difference when we design an MT system?
   In a statistical system probably not. In a rule based system, there may be an effort to make the errors predictable and the output easily post-editable when our aim is dissemination.
3. What is the difference between a minority language and an underresourced language?
   We guess that a minority language is a language with a small number of users. There are underresourced languages which can be considered major, like Indonesian.

Assimilation & dissemination criteria can be used for evaluating the performance of the resulting system. -MK-