Save and reuse human-reviewed translations

A translation memory (advanced tier only) is a repository of human-reviewed translations. Portal users can apply a translation memory on their translation jobs so that exact matches on source-language segments are replaced by their corresponding target-language segments. Use a translation memory to reuse human-reviewed translations and maintain consistency.

You can start with an empty translation memory or import your own translations. As part of the post-editing process, Translation Hub populates your translation memory with the post-edited translations.

Translation Hub keeps only the latest post-edited translations in memory. For example, if a portal user edits a segment that Translation Hub translated with a translation memory, Translation Hub replaces the saved segment with the newly edited segment.

Translation memory compared to other translation resources

In addition to translation memories, you can provide other resources to assist portal users with their translations. The following section describes the differences between each resource.

  • Glossaries match on terms, where as translation memories match on segments. Use glossaries to translate domain-specific terms or entities. Glossaries also has size limits; translation memories can contain as many segments pairs in as many languages as you require.
  • Custom models are trained with your sentence pairs and can help you improve machine translations when you don't have a comprehensive glossary or translation memory available. In cases where you want to tune machine translations for a specific domain and writing style, use AutoML Translation to build custom models that produce better-fitting predictions.

Portal users can use a combination of these resources to help improve the quality of their translations. For more information about how Translation Hub applies resources during translations, see Translate documents.

Before you begin

When you create a translation memory, you can choose to import existing translations. If you do, provide translations in the Translation Memory eXchange (TMX) version 1.4 format. TMX is a standard XML format for providing source and target translations. The following example shows the required TMX structure:

<?xml version='1.0' encoding='utf-8'?>
<!DOCTYPE tmx SYSTEM "tmx14.dtd">
<tmx version="1.4">
  <header segtype="sentence" o-tmf="UTF-8"
  adminlang="en" srclang="en" datatype="PlainText"/>
  <body>
    <tu>
      <tuv xml:lang="en">
        <seg>Hello world!</seg>
      </tuv>
      <tuv xml:lang="es">
        <seg>¡Hola Mundo!</seg>
      </tuv>
    </tu>
  </body>
</tmx>

Segmentation and matching

Segment your source language entries in your TMX file by sentence. During the translation process, Translation Hub segments your content by sentence. Translation Hub then compares these segments with segments in your translation memory. If the segments are an exact match, Translation Hub uses your corresponding translation.

If you include a partial sentence or multiple sentences as a source-language segment, Translation Hub won't match on that segment. As a result, you might have a lower match rate than expected.

Create a translation memory

Administrators create translation memories by using the Google Cloud console.

  1. In the Translation Hub section of the Google Cloud console, go to the Resources page.

    Go to the Resources page

  2. Click Add resource.

  3. In the Add resource pane, select the Translation memories tab.

  4. Specify a name for the translation memory.

  5. Select Create an empty translation memory to create an empty translation memory or select Create a translation memory and import data to import existing translations.

    1. To import existing translations, upload a local TMX file to Cloud Storage or select an existing TMX file from Cloud Storage.

  6. Click Add to create the translation memory.

Add translation memory to portals

After you create a translation memory, add it to an existing portal so that portal users can start using it when requesting translations.

  1. In the Translation Hub section of the Google Cloud console, go to the Resources page.

    Go to the Resources page

  2. From the list of resources, select one or more translation memories to add to one or more portals.

  3. Click Assign to portals, which opens the Assign resource to portal pane.

  4. From the portals field, select one or more portals to add the translation memories to.

  5. Click Assign.

    On the Resources page, you can confirm the addition by viewing the Portal names column for each resource.