Translation Memory is the store of everything Poet has already had translated. When a course is translated, Poet checks the memory first: anything it recognizes is reused automatically, so you don't pay to translate the same sentence twice, and the wording stays consistent from course to course.
You'll find it in the left sidebar, under Translation Memory, alongside Dashboard and Media Library.
What's actually in it?
Two kinds of text, and which of them you see depends on how much you have translated so far.
Poets arrives with the course player's wording already pre-translated into every supported language globally.
If you open Translation Memory before you have translated anything, this is all you will find—a few hundred short strings you don't recognize. That is expected, not an error.
These accumulate as you translate. Every course you put through a translation adds its text to the memory, scoped to your organization. Over time the memory stops being a glossary of button labels and becomes a record of how your organization's compliance language has been translated—reused automatically on every future course.
Scope: Global or your organization
Every entry carries a Scope, and there are exactly two values.
| Scope | Means |
|---|---|
| Global—translated by Confirm | Shared across the whole platform. Mostly player wording and standard phrases. |
| Your organization | Belongs to your organization only. |
There is no course-level scope. An entry either belongs to everyone or to your organization.
Until you have translated a course, every entry you see will be global—the pre-supplied player wording described above.
Finding a specific string
The sidebar page shows a language grid at the ttop—everylanguage with the number of entries held for it—and the entries themselves below, with a search box, a language filter, and a sort.
Select a language tile to filter the entries to that language.
Each entry shows the English source on the left with its character count and the date it was last updated, and the translations on the right, one row per language. Long entries are truncated with a Show full text link.
For anything more involved, use the fuller table at Settings → Translation Settings → Translation Memory. That view adds filters the sidebar page doesn't have:
- Scope — All entries, This organisation, or Global
- Target language and Source language
- Last updated — filter by date
- Sort
It also shows the Scope of every row in a column, which the sidebar page does not.
Fixing a translation that keeps coming out incorrectly/wrong
This is the job Translation Memory is for. If the same phrase is mistranslated across several courses, fix it once here rather than course by course.
- Search for the English source string.
- Check the Scope. If it is Global, see the warning above — raise it with Confirm rather than editing it.
- Select the edit control beside the language you want to correct.
- Replace the wording and save.
The corrected wording is used the next time that string is translated. It does not retrospectively change courses that have already been translated — those keep the wording they were published with. To update an existing course, re-run its translation or edit the segments directly in that course's Translation Workflow.
If a translation is missing entirely
Where a language has no entry for a string, the entry shows + Add a translation. Use it to supply the wording yourself rather than waiting for the next translation job.
Watch the markup
Some source strings carry HTML because the original content was formatted. You will see entries that begin <strong> or <h2 class="ql-align-center">.
Keep the tags exactly as they are. Translate the words between them and leave everything inside the angle brackets alone. A missing closing tag will break the layout of that slide in that language, and it will not be obvious until someone previews the course in Japanese.
Importing and exporting
Both sit at the top of the Translation Memory page.
Export produces the memory as a file—useful for handing your approved terminology to a translation vendor so their work matches what you already have.
Import accepts XLSX or JSON.
- XLSX needs three columns:
sourceText,targetLanguage,translatedText - JSON needs an array of objects with those same three fields
Import is the fast route when you already have an approved glossary, a set of terms the legal team has signed off on, or a wording carried over from a previous supplier. Load it once and every future translation starts from it.
Question
Does an imported translation row overwrite an existing translation for the same source and language, create a duplicate, or get skipped completely?
Answer
It will overwrite the existing translation
Duplicates
The page shows a duplicate count alongside the source-string count. Duplicates are the same source string held more than once—usually because the same sentence arrived from different courses or at different scopes.
They are not harmful, but a high count is a sign the memory would benefit from tidying. If you are seeing inconsistent translations of the same phrase, duplicates are the usual reason.
Who can change it?
Anyone who can reach the Translation Memory page can read it.
Editing individual entries, importing, exporting, and deleting are administrator functions, and the fuller table lives behind Settings → Translation Settings, which is an admin screen.
If you are an author who has spotted a bad translation, the useful thing to do is record the exact English source string and the language and pass it to whoever administers Poet for your organization.
Related
- Translating in Authoring
- Translation Settings—the admin screen behind this page