Translation Memory (TM)

A translation memory (TM) is a database that stores original and translated content for a language pair in translation units (segments). The TM is used to eliminate redundant translation of relevant, already translated segments: translate once, save in TM, reuse as needed.

What is a translation memory good for?

Stored either locally or in the cloud, a TM can be used collaboratively and transferred easily in a compressed file format. It is an ever-growing database that serves as a go-to source for future translations and hence an important economic asset.

Computer-assisted translation (CAT) tools and translation management systems (TMS) with translation memory functionality allow for easy and seamless creation of TMs. Translation memories can be created and categorized according to some of the following parameters:

  • Per language pair,
  • Per topic,
  • Per industry,
  • Per client,
  • Per project,
  • Per user or owner.

Depending on the parameters used, accessibility and usability can be managed as well. This relates to user groups but also to potentially confidential or proprietary information contained in the source.

For example, a project-based translation memory that might contain proprietary content or intelligence can be created based on previous translation memories related to the industry. At the same time, the newly added translation content can be kept subject to the project-specific translation memory alone.

The way the translation memory is created can be managed as well. Since the TM is based on segments, applying specific segmentation rules that serve best for the translation project at hand will be a useful strategy.

How to assert the quality of a translation memory?

Quality assurance of the translation memory is almost as important as its creation and regular updates. Especially in cases where multiple users contribute to a TM, reviewing and correcting the database for best use cases and terminology application according to context is key.

This stems from the fact that in a variety of languages there can be terms that have different meanings depending on the context, thus requiring a different term in the translation. It is also important to keep it as clean and on topic as possible.

Deleting entries from a TM is usually only possible for admins and project managers. Best practices suggest building the asset for consistency and natural-sounding results. To achieve that, guidelines, termbases, and glossaries shared with the creator and user of the TM can go a long way.

The translation memory will only be useful if its quality is maintained on a consistently high level. That will lead to improved results for matching in future translations and provide accurate context to the translators working on any project.

How does a translation memory serve businesses and users?

The translation memory is a data asset. It can be used, re-used, edited, and altered according to certain criteria for best outcomes. It allows its users to not have to translate everything from scratch in situations where source content has been translated previously.

Thanks to its transferability and accessibility, it can be used simultaneously and by multiple users – no matter where those users are. Being updated in real-time, it helps tremendously in complex environments where many linguists and other stakeholders need to contribute to the overall success of each translation project.

The creation, maintenance, and accessibility of a translation memory save project owners and collaborators a huge amount of time, money, and frustration. On top of that, a TM helps to ensure consistency, the natural style of the target language, and it makes it possible to apply commonly-used terminology.