Machine Translation in Industrial Domains: Resources and Evaluations

Resumen

In this work, we describe the ADAPTIA-MT suite of resources for the adaptation and evaluation of Machine Translation (MT) systems in industrial domains. The suite includes specialised terminology and parallel validation corpora for Basque-Spanish translation, manually crafted and validated in four sectors: automotive, energy, railways and machine tool. We build upon these resources to compare two main approaches on domain adaptation tasks, namely Neural Machine Translation and MT based on Large Language Models, measuring both general translation quality in the selected domains and terminological accuracy. The resources are shared with scientific community under a CC-BY-NC-ND 4.0 license.