Developing Prototypes for Machine Translation between Two Sami Languages

Proceedings of the 13th Annual Conference of the EAMT, pages 120–127, Barcelona, May 2009 Developing Prototypes for Machine Translation between Two Sámi Languages Francis M. Tyers Departament de Llenguatges i Sistemes Informàtics, Universitat d’Alacant E-03071 Alacant (Spain) ftyers@dlsi.ua.es Linda Wiechetek Giellatekno, Universitetet i Tromsø, Norway linda.wiechetek@uit.no Trond Trosterud Giellatekno, Universitetet i Tromsø, Norway trond.trosterud@uit.no Abstract This paper describes the development of two prototype systems for machine trans- lation between North Sámi and Lule Sámi. Experiments were conducted in rule-based machine translation (RBMT), using the Apertium platform, and statistical ma- chine translation (SMT) using the Moses- decoder. The experiments show that both approaches have their advantages and dis- advantages, and that they can both make use of pre-existing linguistic resources. 1 Introduction In this paper we describe the development of two prototype machine translation systems between two Sámi languages, North Sámi (sme) and Lule Sámi (smj), one rule-based (Apertium), and one statistical (Moses). There are other systems which have been developed with marginalised languages in mind (e.g. (Lavie, 2008)), but, as of writing, these were not available under an open-source li- cence and thus could not be applied to the task at hand. The content will be split into several sec- tions. The first section will give a general overview of the languages in question, and sketch a typol- ogy of MT scenarios for minority languages. The next sections will describe the two machine trans- lation strategies in some detail and will outline how the existing language technology was able to be re-used and integrated. We will follow this by a short evaluation and then some discussion and future work. 1.1 The languages Both North Sámi and Lule Sámi belong to the Finno-Ugric language family and are spoken in the c⃝ 2009 European Association for Machine Translation. north of Norway and Sweden, North Sámi also in Finland. North Sámi has between 15,000 and 25,000 speakers, while Lule Sámi has less than 2,000 speakers. The Sámi proto-language was originally an ag- glutinative language, but North and Lule Sámi have developed features known from inflective lan- guages (case/number combinations are often ex- pressed by one suffix only, certain morphologi- cal distinctions are expressed by means of conso- nant gradation (i.e. a non-segmental process) only, etc.). The main objective with the development of the prototype rule-based system was to evaluate how well existing resources could be re-used, and if the shallow-transfer approach was suited to languages with more agglutinative typologies. 1.2 A typology of MT systems for minority languages Minority language speakers typically differ from the majority in being bilingual, the minority speaks the language of the majority, but not vice-versa. This has some implications for the requirements society will put to machine translation systems. A majority to minority language system must be of high quality, so high that post-editing the output is faster than translating from scratch. The goal is to produce well-formed text, not to understand the content, since the minority language users will pre- fer the original to a bad translation. …

Content truncated to 500 words