Quality Estimation of Machine Translation for Low-Resource Languages

Organisatsiooni nimi
TartuNLP
Kokkuvõte
The goal of this thesis is to apply the mechanistic interpretability of neural networks to automatic prediction of translation quality.
Our team works on machine translation for low-resource Finno-Ugric languages. To evaluate the quality of the translation during a training phase, we compare the translations with human references. However, for the online system, the evaluation should be done in real time, without relying on human references.
Lõputöö kaitsmise aasta
2024-2025
Juhendaja
Lisa Yankovskaya, Mark Fišel
Suhtlemiskeel(ed)
inglise keel
Nõuded kandideerijale
Tase
Magister
Märksõnad
#machine_translation #quality #neural_networks

Kandideerimise kontakt

 
Nimi
Lisa Yankovsksya
Tel
E-mail
lisa.yankovskaya@ut.ee