Agreement with AT-LanguageSolutions in pro of a R&D project on neural machine translation models.
This collaboration aims to develop new machine translation systems based on Deep Learning. The systems based on deep learning use machine learning algorithms, basically multi-layer neural networks, capable of learning from already translated texts.
The advantage of these algorithms with regards to previous machine translation systems, such as statistical systems, is that they are capable of generalising morphological elements and syntactic structures that were not present in the training material (translated texts from which the system has learnt).
The improvement in these systems is having a major impact on the scientific and business community.