Please use this identifier to cite or link to this item:

http://hdl.handle.net/10609/97446
Title: Evaluación de traducciones realizadas con un modelo neuronal y uno estadístico: valoración de resultados para los pares francés-español e inglés-español
Author: Crosby Casali, Cristina
Tutor: Oliver González, Antoni
Keywords: statistical machine translation
neural machine translation
augmented translation
Issue Date: 17-Jun-2019
Publisher: Universitat Oberta de Catalunya (UOC)
Abstract: This project compares statistical and neural machine translation systems, and assesses translations generated by models of both systems created on open source frameworks. The first model is created using Moses, a structure that allows the creation of statistical machine translation (SMT) models. While the second model is created using Marian, a toolkit that supports the creation of neural machine translation (NMT) models. MultiUN bilingual corpora are used, which allow to build both SMT and NMT models with quantity and quality of data for the French-Spanish and English-Spanish translations. The resulting translations are assessed using manual metrics, including a light post-editing process and error annotation. The results are benchmarked with a third translation obtained from an open-access machine translation engine. Automatic metrics are calculated with the three translations. The evaluation helps to define the strengths and weaknesses of each system, and to identify the most common translation errors. The performance of the models allows to ascertain if they could be considered an integral part of a broader process of augmented translation, which regards machine translation as a tool, while the translator is at the core of the translation workflow.
Language: Spanish
URI: http://hdl.handle.net/10609/97446
Appears in Collections:Bachelor thesis, research projects, etc.

Share:
Export:
Files in This Item:
File Description SizeFormat 
ccrosbyTFM0619memoria.pdfMemoria del TFM1.42 MBAdobe PDFView/Open

This item is licensed under a Creative Commons License Creative Commons