Building Neuronal Mcahine Translation methods and systems to improve coherence at paragraph and document level
(2020 - 2021)
After the significant progress of neuronal machine translation, one of the main current challenges in the field is to increase the coherence of TA systems in the translation of whole paragraphs or documents. The current systems translate at phrase level, without any access to global context. This limitation produces systematic errors that are undertaken in this project for the es-eu and fr-eu.language pairs.
Organization: Eusko Jaurlaritza
Main researcher: Kepa Sarasola
Iñaki Alegria, Olatz Ansa, Nora Aranberri, Jose Mari Arriola, Gorka Labaka, Mikel Lersundi, Olatz Perez de Viñaspre, Kepa Sarasola, Xabier Soto, Ruben Urizar