Skip to content
English
Submit a request
  • There are no suggestions because the search field is empty.

What Are Translation Memories (TMs)?

Translation Memories (TMs) are linguistic databases that store previously translated segments, typically sentences or short paragraphs, together with their source text. Whenever the same or a similar segment appears in new content, the stored translation can be reused automatically.

How Lara Supports TMs

Lara supports translation memories natively: you can create, manage, and leverage your own TMs directly within the platform. Your previously translated content is reused in future translations, ensuring consistency and efficiency across all your projects.

For a deeper dive, check out our articles on Using Translation Memories in Lara’s UI and Add to Memory: Build Your Own Translation Memory.

How Do Translation Memories Work?

When a translator works on a text in a CAT tool (Computer-Assisted Translation tool) such as Matecat, Trados, memoQ, or Smartcat, each pair of source and target segments is saved in the TM.

When a new project contains a segment that already exists in the TM, the tool suggests the saved translation:

  • Exact match (100%): the new segment is identical to one already stored.
  • Fuzzy match: the new segment is similar but not identical (for example, only a word or a number has changed). The tool suggests the closest translation, so the translator only needs to adjust the differences.

What Is a TMX File?

TMX (Translation Memory eXchange) is the standard, XML-based format used to store and exchange translation memories between different tools. Thanks to TMX, you can export a TM from one CAT tool and import it into another, including Lara, without losing your linguistic assets.

Here’s what a simple TMX file looks like:

<?xml version="1.0" encoding="UTF-8"?>
<tmx version="1.4">
  <header creationtool="ExampleTool" creationtoolversion="1.0"
          segtype="sentence" o-tmf="TMX" adminlang="en-US"
          srclang="en-US" datatype="plaintext"/>
  <body>
    <tu>
      <tuv xml:lang="en-US">
        <seg>Click Save to apply your changes.</seg>
      </tuv>
      <tuv xml:lang="it-IT">
        <seg>Fai clic su Salva per applicare le modifiche.</seg>
      </tuv>
    </tu>
    <tu>
      <tuv xml:lang="en-US">
        <seg>Your subscription will renew automatically.</seg>
      </tuv>
      <tuv xml:lang="it-IT">
        <seg>Il tuo abbonamento si rinnoverà automaticamente.</seg>
      </tuv>
    </tu>
  </body>
</tmx>

The main elements are:

  • <header>: general information about the file, such as the source language (srclang) and the tool that created it.
  • <tu> (translation unit): a single source–target pair.
  • <tuv> (translation unit variant): the version of the segment in one language, identified by the xml:lang attribute.
  • <seg>: the actual text of the segment.

A TMX file can contain thousands of translation units and, in some cases, more than two languages per unit.

What Are the Main Benefits?

  • Terminology consistency: repeated content is always translated the same way.
  • Faster turnaround: segments that have already been translated don’t need to be translated again.
  • Stable quality: especially valuable for long-term projects and brand-specific language.
  • Reusable assets: thanks to the TMX standard, your TMs aren’t locked into a single tool.

Who Uses Translation Memories?

TMs are especially useful for:

  • Companies with recurring content, such as product descriptions, manuals, or legal documents
  • Translation teams working on ongoing or large-scale projects
  • Language service providers managing multilingual workflows

Are Translation Memories the Same as Machine Translation?

No. A Translation Memory is built from previously approved translations, usually produced or reviewed by humans, while Machine Translation (MT) generates new translations in real time using AI.

The two technologies are often used together. In post-editing workflows, for example, TM matches are prioritized and MT fills in the gaps when no match is available.

Are TMs Private or Shared?

TMs can be:

  • Private: linked to a single translator, project, or client.
  • Shared: used across teams, groups, departments, or agencies to maintain consistent language at scale.

This article is about:

  • What Translation Memories (TMs) are
  • How TMs work in translation workflows, including exact and fuzzy matches
  • What a TMX file is and how it’s structured
  • The benefits of using TMs (consistency, speed, quality)
  • The role of TMs in hybrid workflows combining AI and human post-editing
  • How Lara supports TMs, allowing you to build and manage your translation memories directly in the platform