What is the memory auto-clean feature, and how does it work?

Affected users

  • All Memory users

Wezen version

  • Version 11.0 and above
  • Migrated memory only

Auto-cleaning feature

In order to avoid having too many versions of a translation, for a language, saved into the memory, Wezen implemented the auto-cleaning feature, in order to have a cleaner and more efficient translation memory.

Wezen's translations are ranked from the most to the least relevant (We prioritize the latest translation validated at the highest step of the workflow). As for now, we accept only 10 translations for a source segment. It means that when a translation goes into the memory and is more relevant than an existing one, it pushes down the translations and the 11th one is automatically deleted.

Here is an example on how the auto-clean works

In this memory, it already has 10 translations in en-US saved for the source segment in fr-FR. We have an old translation (highlighted in blue) that has the oldest last usage time at the highest step.

Memory widget interface:

Memory interface:

Once a new translation entry is being made during the translation process, for example, Wezen will determine if the newest entry is relevant when compared to the rest of existing translations.

If it's relevant and is a completely new entry (i.e this translation version never existed before), Wezen will classify and rank it based on the translation's relevancy, save it into the memory, and remove the least and oldest translation automatically.

Based on the example below, a user validated a new translation, and since it's more relevant, it is saved into the memory, and the least relevant and oldest translation is automatically removed.

Memory widget interface:

Memory interface:

In the case when the newest entry is very irrelevant, after being compared to the rest of the existing translations, it will not be considered for being saved into the memory.