Files, Files Everywhere: The Subtle Power of Translation Alignment

Ciklopea 6 years ago 5 min read

Here’s the basic scenario: you have the translated versions of your documents, but the translation wasn’t performed in a CAT tool and you have to build a translation memory because these documents need to be updated or changed across the languages, you want to retain the existing elements, style and terminology, and you have integrated CAT technology in your processes in the meantime. The solution is a neat piece of language engineering called translation alignment.

Translation alignment is a native feature of most productivity tools for computer-assisted translation, but its application in real life is limited to very specific situations, so even the language professionals rarely have an opportunity to use it. However, these situations do happen once in while and when they do, alignment usually comes as a trusty solution for process optimization. We will take a look at two actual cases to show you what exactly it does.

Example No. 1: A simple case

Project outline:

Three Word documents previously translated to one language, totaling 6000 unweighted words. Two new documents totaling around 2500 words that feature certain elements of the existing files and need to follow the existing style and terminology.

Project execution:

Since the translated documents were properly formatted and there were no layout issues, the alignment process was completed almost instantly. The software was able to segmentize the source files and we matched the translated segments, with some minor tweaking of segmentation. We then built a translation memory from those matched segments and added the new files to the project.

The result:

Thanks to the created translation assets, the final wordcount of the new content was around 1500 and our linguists were able to produce translation in accordance with the previously established style and terminology. The assets were preserved for use on future projects.

Translation alignment in action.

Example No.2: An extreme case of multilingual alignment

Project outline:

In one of our projects we had to develop translation assets in four language pairs, totaling roughly 30k words per language. The source materials were expanded with new content totaling about 20k words unweighted and the language assets had to be developed both to retain the existing style and terminology solution and to help the client switch to a new CAT platform.

Project execution:

Unfortunately, there was no workaround for ploughing through dozens of files, but once we organized the materials we could proceed to the alignment phase. Since these files were localized and some parts were even transcreated to match the target cultures, which also included changes in layout and differences in content, we knew that alignment was not going to be fully automated.

This is why native linguists in these languages performed the translation alignment and communicated with the client and the content producer during this phase. While this slowed the process a bit, it ultimately yielded the best results possible.

We then exported the created translation memory in the cross-platform TMX format that allowed use in different CAT tools and the alignment phase was finished.

The result:

With the TM applied, the weighted volume of new content was around 7k words. Our linguists localized the new materials in accordance with the existing conventions in the new CAT platform and the translation assets were saved for future use.

Wrap up

In both cases, translation alignment enabled us to reduce the volume of the new content for translation and localization and ensure stylistic and lexical consistency with the previously translated materials. It also provided an additional, real-time quality control and helped our linguists produce a better translation in less time.

Translation alignment is not an everyday operation, but it is good to know that when it is called to deliver the goods, this is exactly what it does.

Like This Article? Subscribe to Receive More Via Email

  • receive a digest with new articles
  • up to 2 emails a month

Related Articles

Maximizing Translation Efficiency with AI-Powered TM Cleanup: The Future of Localization

2 months ago

In the fast-paced world of localization, maintaining the accuracy and quality of Translation Memories (TMs) is essential for ensuring seamless, consistent translations. As companies scale globally, the size of their TMs grows exponentially, which can lead to inconsistencies, outdated translations, and even redundant content. That’s where AI-powered TM Cleanup comes into play. This innovative solution combines the strengths of AI with prompt engineering, revolutionizing how organizations manage their TMs.

Continue reading