Machine translation in a professional translation workflow: post-editing or computer aided translation
Localization is a fascinating field which has changed dramatically in the last years and it is still changing and adapting itself as society is. The use of machine translation has developed this field astonishingly. Many companies have started including machine translation in their workflows to meet...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis doctoral |
| Estado: | Versión publicada |
| Fecha de publicación: | 2023 |
| País: | España |
| Institución: | CBUC, CESCA |
| Repositorio: | TDR. Tesis Doctorales en Red |
| OAI Identifier: | oai:www.tdx.cat:10803/687972 |
| Acceso en línea: | http://hdl.handle.net/10803/687972 |
| Access Level: | acceso abierto |
| Palabra clave: | Machine translation Traducción automática Traducció automàtica 81 |
| Sumario: | Localization is a fascinating field which has changed dramatically in the last years and it is still changing and adapting itself as society is. The use of machine translation has developed this field astonishingly. Many companies have started including machine translation in their workflows to meet their tight deadlines and be more competitive in terms of budget. The aim of this work is to analyze several significant issues that must be considered when it comes to implementing machine translation in a professional translation workflow. We have carried out an experiment in which real texts from a company have been translated using human translation, statistical machine translation and neural machine translation. After that, a mixed pool of 40 experienced and novice post-editors has post-edited these texts. We have analyzed quantitatively the speed and edit distance of the post-editing phase and at the same time we have reviewed the quality of the different translations. Finally, we have used a qualitative approach in the shape of a questionnaire of 6 questions that post-editors have answered. This experiment with 40 post-editors has shown that, in a practical setup, human translation (HT) and machine translation (MT) are not that far away each other. Our research confirms that HT and MT can be interrelated. In addition to that, our work has confirmed that some post-editors think HT is MT, and the other way around. This means that MT allows post-editors to work as naturally as if they were reviewing HT. The results of the experiment and the feedback from post-editors also confirm that Neural Machine Translation (NMT) has helped to close the gap between HT and MT. Last but not least, results in our experiment also confirm that post-editing (PE) is more competitive than HT from an economic point of view. |
|---|