Case Study: The Backbone of Machine Translation Post-Editing (MTPE)

At Idea Translations, we are proud to be at the forefront of industry innovations. We constantly adapt to
technology trends to provide our clients with top-notch, high-quality services using the latest tech. That’s
why we’ve included Artificial Intelligence into selected translation projects.

While AI plays a helpful role in the initial stage of some translation projects, our linguistic professionals
then take over to make the necessary adjustments and ensure the cultural context, idioms, and client terminology are correctly captured. AI helps us save time and costs. However, on its own, it cannot meet the high standards our clients expect.

That’s why, at Idea Translations, every translation undergoes a strict quality control and review process. In
this Case Study, we reveal the step-by-step procedure involved in Machine Translation Post-Editing
(MTPE).

Join us as we demonstrate the fundamental process behind MTPE. Keep in mind that additional steps, such as Editing, Proofreading, and Desktop Publishing, often come into play for most projects, and our team collaborates closely with subject matter experts for successful project completion.

MTPE PROCESS

Step 1: Project Creation and Pre-Translation

We get started on our translation project by using the most appropriate CAT tool. In this case, we are using Trados. We proceed to creating the project, selecting the language pairs, and chosing the desired file path. Then, in the Translation Resources tab, we implement the most suitable AI engine.

In this case, we are going with the Systran API. Lastly, we upload our client’s Translation Memory (TM) and Term Base (TB) to ensure existing matches are recognized. In no time, our file is pre-translated and ready for easy post-editing.

Step 2: Post-Editing

Expert linguists meticulously review and enhance AI-generated translations. They ensure precise conveyance of meaning, resolve formatting issues such as double spaces, and maintain consistency in terminology.

Step 3: Quality Assurance (QA)

We do an additional check of the translation with our QA tools, such as XBench. This software easily identifies issues such as identical source and target segments, terminology mismatches, wrong numbering and spacing. Then, we review the Xbench report and correct any errors in the bilingual file. Once this is done, we transfer the reviewed translation to a text processing software for a thorough spellcheck. Finally, we generate the final target file.

Step 4: Formatting Verification

Before the final file is ready, we carefully review it to ensure it closely matches the original document in terms of layout. We fix any design issues and perform a final spellcheck for added assurance (just in case!).

CONCLUSION:

This Case Study serves as evidence that relying solely on Artificial Intelligence in the translation process falls short, as it lacks the indispensable human elements of common sense, context, and sentiment. As a result, we believe that a combination of both AI and human expertise is indispensable to maximize efficiency and enhance our resource utilization while guaranteeing an end product of the highest quality.