We started researching Machine Translation in 2000 and have followed the paradigms being developed in the area: first RBMT, then SMT and currently NMT. We have focused mainly on translation from and into Basque since, in addition to its commercial interest in our country, it is an important challenge for several reasons: the complexity of Basque morphology, the free order of sentence constituents, and the scarcity of resources. The results have been very good and the tools developed are being used commercially. Matxin, OpenTrad and Modela are the well-known brands that represent our technology.

In addition to the good results for Basque, we have collaborated with other groups and companies in order to improve machine translation for other languages (English, French, Quechua...).

In recent years, we have also tackled unsupervised machine translation, achieving an important scientific impact (reference in Science journal).