(w j. angielskim)
Case Study: Machine Translation + Post-Editing for Agence Europe’s Daily Bulletin
Agence Europe, a Brussels-based news agency focused on European economic and political integration since 1953, was struggling under the challenge of publishing its daily news in two languages on the same day. News is news when it is new, meaning that delays with translation were out of the question. This kept the client’s team of linguists and their coordinator under constant time pressure.
The news articles are written in French and delivered by journalists throughout the day to an online editorial tool designed by the client. The coordinator distributed the news articles for translation into English among the linguists via the editorial tool. Due to the time restraints, the translations were merely revised by a proofreader. They were done directly in the client’s editorial tool, which is not equipped with a translation memory, terminology database, quality assurance tools or even a specific spellcheck, and this was putting the translation quality at risk.
Agence Europe was looking for a solution to reduce the stress on its team and increase the translation quality without extra expense, so they partnered up with Translavic. Given that considerable volumes of similar content type (news articles on EU affairs) need to be translated within a few hours, five days per week, while maintaining top quality, Translavic assessed this project as suitable for a trial with machine translation (MT), with full post-editing.
After running a few human evaluation tests on MT output from various providers, the MT provider with the highest fluency was chosen.
To ensure top quality translations, there was a strict selection of post-editors, who were trained specifically for the project. Post-editing of machine translation is a service that is not widely adopted yet among professionals and buyers in the translation industry. Thus, it is of utmost importance to agree with the client on the expected quality standard, and to instruct the team of linguists accordingly. For this project, top-notch publishable quality is required, so the team of post-editors was instructed to use as much MT output as possible, while implementing all corrections necessary to ensure that the end product is equal to human translation in terms of fluency, accuracy, appropriate and consistent style and terminology. In agreement with the client and linguists, a Style Guide was compiled for the project. To ensure the correct use of terminology, a terminology database was created and is constantly updated by a terminologist, with the client’s approval. The terminology database is also updated with preferred or forbidden terms according to feedback received from the client.
To reduce coordination time and to ensure quick and easy file exchange, Translavic looked into ways of automating the process and opted for a cloud-based tool allowing full automation of the following steps:
- importing the articles from the client’s editorial tool to create a translation project using the template that includes the translation memory, term base and a pre-defined pool of dedicated post-editors;
- pre-translation with the translation memory and machine translation;
- distribution of work among the pool of post-editors;
- exporting the final post-edited articles to the client’s editorial tool.
This way, the time needed for the coordination of the project was considerably reduced, freeing the dedicated project manager from routine tasks, to be able to focus on communication with the client and the team of post-editors, handling feedback and any after-delivery modifications.
The team of post-editors works in a cloud-based tool (monitored live by the project manager) that has a spellcheck and allows a thorough verification of formatting tags, punctuation, numbers, consistency and terminology. After the post-editing step, a final check by the project manager is performed, where a final verification of the completeness and compliance with the style guide and client’s instructions is carried out.
Such workflow ensures a comfortable environment for the team to work in. Freed from the mundane tasks, which are taken over by machine translation and process automation, the team can focus on more creative tasks such as finding the right equivalent in the target language for a metaphor used in the source article, or keeping the communication flow within the team and with the client smooth and sound.