The FISKMÖ Project: Resources and Tools for Finnish-Swedish Machine Translation and Cross-Linguistic Research

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 3808–3815 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 3808 The FISKM ¨O Project: Resources and Tools for Finnish-Swedish Machine Translation and Cross-Linguistic Research J¨org Tiedemann∗, Tommi Nieminen, Mikko Aulamo∗, Jenna Kanerva†, Akseli Leino†, Filip Ginter†, Niko Papula⋄ ∗Department of Digital Humanities, †Department of Future Technologies, ⋄Multilizer University of Helsinki, University of Turku, Espoo {jorg.tiedemann, mikko.aulamo}@helsinki.fi, {jmnybl,akeele,figint}@utu.fi, niko.papula@multilizer.com Abstract This paper presents FISKM ¨O, a project that focuses on the development of resources and tools for cross-linguistic research and machine translation between Finnish and Swedish. The goal of the project is the compilation of a massive parallel corpus out of translated material collected from web sources, public and private organisations and language service providers in Finland with its two official languages. The project also aims at the development of open and freely accessible translation services for those two languages for the general purpose and for domain-specific use. We have released new data sets with over 3 million translation units, a benchmark test set for MT development, pre-trained neural MT models with high coverage and competitive performance and a self-contained MT plugin for a popular CAT tool. The latter enables offline translation without dependencies on external services making it possible to work with highly sensitive data without compromising security concerns. Keywords: parallel corpora, machine translation 1. Introduction Finland is a country with two official languages, Finnish and Swedish. The demand for translation between the two languages is huge and translation efforts produce a lot of overhead and costs in public administration and organisa- tions as well as any service providers in the country. Yet there is no systematic collection of translated data nor any large scale effort to develop translation services and tools for the public and administrative use in the country. For this reason, we started FISKM ¨O1 (finsk-svensk korpus och maskinovers¨attning) a joint project of the universities of Helsinki and Turku and Kites (an umbrella organization for the language sector in Finland), currently funded by the Swedish Culture Foundation.2 The primary goal of the project is to create a massive par- allel corpus of Finnish and Swedish to support linguistic research and machine translation (MT) development. The secondary goal of the project is to establish a public ma- chine translation service for translation between Finnish and Swedish. As the quantity and quality of training data is the most important factor in machine translation quality, the FISKM ¨O translation service can be expected to pro- vide translations of higher quality than other public ma- chine translation services, for which less training data is available. Our mission is to systematically collect material that has been translated from Finnish to Swedish and vice versa in the public sector, in private organisations and at language service providers in Finland. We intend to publish the data with permissive licenses to increase the impact on on-going research and development. Nevertheless, we will also in- clude restricted data …

Content truncated to 500 words