Ulike veger fra nordsamisk til lulesamisk

Abstract

The present article evaluates different approaches to the construc-tion of a North Saami-Lule Saami dictionary. The pivot method (via Norwegian) gives rise both to a low coverage and to a massive overgeneration and requires other methods in order to distinguish between good and bad candidate pairs. We show that it is possible to distinguish reliably between good and bad pairs without addi-tional semantic resources, only by using grammatical analysers, a transliteration component and a grammatical machine translation (MT) system. Similar resources are available also for other Saami languages, and the results shown here are likely to be portable also to other similar language pairs.