The paz-andrade law and the use of portuguese for automated translation in galego

In order to obtain high-quality corpus-based, statistical and neural machine translation, large volumes of human translations are needed between the two languages. Some low-resource languages do not have this volume of resources, but resources may be accessible from close-related languages or variants. Nevertheless, it is difficult to develop machine translation systems that are officially recognized by the language authorities of the language with fewer resources if there are sociolinguistic controversies with respect to their relatedness, further strengthened by legal frameworks. In this article, we will see how a law that untangles the controversial Galician-Portuguese linguistic relations (i.e., Paz -Andrade Law) legally facilitates the development of neural machine translation (NMT) systems with good results for Galician, thanks to the use of Portuguese corpora. Finally, based on this experience, we propose a method to develop neural translation systems that can be applied in the context of other low-resource languages that, despite being controversial, have a close relation with a language with greater resources.