Error Analysis and Solutions of Translation Agencies Grata and Interlex Machine Translation Models
Name
Karen Saksakulm
Abstract
Machine translation is a machine learning sector that is gaining popularity. It simplifies handmade translations and helps to save time. Several model architectures alongside pre- and post-processing methods have been introduced, but a single most effective translating solution has never been found. During the research of this thesis, the data of two translation companies, Grata and Interlex, was used for pre- and post-processing, model training and translation processes. The goal was to compare the translation accuracy of different machine translation models, analyze encountered problems and propose solutions for fixing these issues. The results showed that one of the models is incapable of translating the input data.
Graduation Thesis language
Estonian
Graduation Thesis type
Bachelor - Computer Science
Supervisor(s)
Andre Tättar, Liisa Rätsep
Defence year
2021