At the Conference on Machine Translation, taking place on December 7-8, 2022 in Abu Dhabi (UAE), the results of the General MT Shared Task competition were officially announced, in which a team from the Faculty of Mathematics and Computer Science at UAM, took first place! Artur Nowakowski, Gabriela Palka, Mikolaj Pokrywka, Kamil Gutmann designed a translation model that performs better than those of competitors such as Google, Microsoft, Yandex and Amazon.
Artur Nowakowski and Gabriela Palka (PhD students from WMI) and Mikołaj Pokrywka and Kamil Guttmann (students majoring in Computer Science, participating in the AITech project) designed a machine translation model that achieved better efficiency than all other submitted models, losing only to human translation.
This year's General MT Shared Task was to create the best possible translator capable of translating texts from various fields. They beat competitors such as Google, Microsoft, Yandex and Amazon, as well as teams from Charles University in Prague, Huawei and INRIA (French National Institute for Research in Digital Science and Technology).
This year, news, social media, human conversation (e.g., chat) and e-commerce were the main areas of focus for the competition. Our representatives took part in Czech-Ukrainian and Ukrainian-Czech translation. It's worth mentioning that they used only the data provided by the competition organizers (marked with a white background in the results table) to create the system, unlike many teams using their own data in addition. The creation of the translation system took about two months. During that time, they tried various methods based on neural networks that would be able to achieve the best possible translation. Designing such an efficient system would not have been possible without the powerful computing apparatus of the UAM Artificial Intelligence Center.
