Grata Continues Cooperation with the University of Tartu
Grata continues its cooperation with the University of Tartu to develop a subject field specific machine translation engine. The results of the pilot project – carried out in 2016 to evaluate the potential of the development idea – showed that the translated output of such an engine is good enough to try post-editing with it and that further enhancement and tuning of the engine will most likely lead to desired results. Therefore, Grata will continue to cooperate with the University of Tartu during a follow-up project running from April 2017 to January 2018.
The purpose of the project is to create a subject field specific machine translation engine for Grata, using Grata’s own databases and translation memories as the basis, for translating technical texts from English and German into Estonian. The desired outcome of the project is a customized machine translation engine that can be integrated into our workflow – this means that the quality of the translated output has to be so good that it would take less time to post-edit it than to translate the text from scratch.
Over the last few years we have started to receive more and more inquiries about post-editing services since this is the direction the translation industry is heading in. As a synthetic and grammatically complex language, Estonian has struggled somewhat more to keep up with this trend than some more analytic languages. Grata believes that in order to start providing proper post-editing services for Estonian in the future, it is necessary to cooperate with translation specialists as well as technical experts to develop a translation engine that has been optimized and customized according to the needs, language pairs and specialization of the translation agency. Translation agencies cannot use widely spread and general translation engines as this would result in quality problems and confidentiality breaches. A reliable translation engine would increase translation speed of certain types of texts, which means bigger volumes with shorter deadlines and for a better price for the client and higher productivity for the translation agency.
Recent developments in the Estonian machine translation research have shifted the focus of Grata’s project – the pilot project was carried out using statistical machine translation, but recent experiments conducted by the University of Tartu have shown that neural machine translation would give even better results. Therefore, neural approach will be adopted for the follow-up project.
The project will increase the technical competence and awareness of the company, improve employees’ knowledge of machine translation and its principles, and increase the number of translators with post-editing skills on the Estonian translation market. The project also contributes to the research and development of Estonian machine translation in general and promotes cooperation between translation agencies and research institutions.
The European Regional Development Fund finances the project with 12,902.40 EUR.