Translation Memory (PRO)

Translation Memory is a feature that automatically saves every segment you translate and reuses those translations when the same or similar text appears elsewhere on your site. Think of it as a personal translation database that grows smarter the more you use Lang Forge. Over time, Translation Memory dramatically reduces repetitive work, improves consistency across your site, and even saves AI credits by reusing stored translations instead of generating new ones.

Translation Memory is especially powerful for sites with recurring content patterns — product descriptions that follow a template, legal pages with standard paragraphs, support articles with common phrases, FAQ answers that repeat across multiple pages, and any content where the same sentences appear in different contexts.

How Translation Memory works behind the scenes

Every time you translate a segment — whether you type it manually in the Visual Editor, edit it in the post editor, accept an AI-generated translation, or click an AI re-generate button on a single segment — Lang Forge stores the original text and its translation as a matched pair. This happens automatically. You do not need to take any extra steps or “save to memory” — it simply builds itself as you work. Single-segment AI calls (the per-field “AI Translate” button, the Visual Editor’s per-segment re-generate, the Frontend Visual Editor inline edit, and SEO meta translations done by the SEO Forge bridge) all feed the same TM table — so the memory grows whether you translate a whole post or fix one sentence at a time.

The next time you encounter a segment that matches (or is similar to) a previously translated one, Translation Memory offers the stored translation as a suggestion:

How fuzzy matching works in practice

Fuzzy matching compares text at the word level, not character level. This means it handles variations intelligently. For example:

Where Translation Memory suggestions appear

• Visual Editor: When you click a segment to translate, a suggestion panel appears below the field showing any TM matches. Click a suggestion to insert it, then edit if needed
• AI Translation: When the AI translates content, it checks Translation Memory first. If an exact match exists, it uses the stored translation instead of generating a new one. This saves AI credits and ensures consistency
• Bulk Translation: During site-wide AI translation, TM matches are applied automatically across all posts. Posts that heavily reuse content from other already-translated posts will use very few AI credits

Managing Translation Memory in the admin

Open Lang Forge → Translation Memory. The page lists every pair LangForge has captured, paginated 25 entries per page. From here you can:

• Search by source or target text (the box at the top of the toolbar) — debounced 250 ms, matches anywhere in the segment.
• Filter by language pair with the From / To dropdowns. Leave both as “Any …” to see the full memory.
• Edit a pair: click the pencil icon on any row. A modal opens with the source (read-only — it’s the unique key) and target (editable) text. Save to update the stored translation; future Visual Editor suggestions and AI exact-match lookups use the new value immediately.
• Delete a pair: click the trash icon. A confirm dialog warns you the deletion is permanent.
• Populate from existing translations: the toolbar button walks every translation group on the site and writes one TM pair per matched paragraph. Use this once after migrating from WPML, Polylang, or TranslatePress.
• Import a TMX file (toolbar → Import TMX): drag-drop a TMX 1.4 file (.tmx or .xml, max 10 MB). After upload you’ll see the number of entries imported, the number of duplicates skipped, and any parse errors. Tick Overwrite existing entries if you want incoming pairs to replace any same-source pair already in memory; otherwise duplicates are skipped.
• Export TMX: toolbar → Export TMX downloads the entire memory as a TMX 1.4 XML file. Filter by From / To first to scope the export to a specific language pair (filename: langforge-tm-{from}-{to}-YYYY-MM-DD.tmx).

The empty state (“Your Translation Memory is empty”) appears when no pairs exist yet. From there a single click on Populate from existing translations seeds the memory if your site already has translated posts.

If you only need to inspect or migrate stored pairs without using the UI, you can also reach the same data through:

• Lang Forge → Tools → Populate Translation Memory (alternate entry point for the populate action).
• Lang Forge → Import / Export for XLIFF / CSV round-trips between sites — TM rebuilds itself from those imports as the translations are saved.
• Direct wp_lf_translation_memory SQL for power users (see Developer Guide).

Real-world example: Translation Memory saving time on a product catalog

An online store sells 300 products. Many product descriptions share common phrases: “Free shipping on orders over $50”, “30-day money-back guarantee”, “Available in multiple sizes and colors”, “Made from premium materials”. The content manager translates the first 50 products into Spanish. Translation Memory stores all these common phrases. When she moves to products 51 through 300, TM suggests exact matches for the recurring phrases — she accepts them with one click instead of retyping them. For a catalog of 300 products, Translation Memory saves an estimated 4-6 hours of translation work and ensures that “30-day money-back guarantee” is translated identically on every single product page.

What happens on the Free plan

Translation Memory is not active on the Free plan: no translation pairs are stored, and the Visual Editor’s TM suggestion panel does not appear. If you upgrade to PRO later, TM starts accumulating from that point forward. It does not retroactively index translations made before the upgrade, but any new translations (including edits to existing ones) are saved into TM. To backfill TM from translations that existed before you turned PRO on, use Lang Forge → Tools → Populate Translation Memory — it walks the existing translation groups and seeds pairs into the TM table.

Common mistakes to avoid

• Treating TM as a once-and-done list. TM grows automatically as you save translations. Bad pairs come from bad source translations — fix those at the source post or use the Translation Memory page’s edit modal directly to repair the cached pair, then save
• Skipping TM populate after migration. When you migrate from WPML, Polylang, or TranslatePress, the existing translations don’t auto-seed TM. Click Populate from existing translations on the Translation Memory page once after migration so AI translations of new content start picking up the in-place translations as exact matches
• Importing a TMX without checking the Overwrite toggle. By default, duplicate source segments are skipped on TMX import. If your TMX has the freshest, most-correct copy of every translation, tick Overwrite existing entries before clicking Import — otherwise older pairs already in memory win

> Tip: Translation Memory is most valuable for sites with standardized content. If your site uses templates or patterns (product descriptions, job listings, property listings, course descriptions), TM will quickly learn the common elements and save enormous amounts of time.

> Good to know: TMX export and import use the standard TMX 1.4 format, so any CAT tool (memoQ, Trados, OmegaT, Smartcat, …) can swap memories with LangForge directly. For an agency you’d ship a TMX from Lang Forge → Translation Memory → Export TMX; for a translation request returning a TMX, drop it into Import TMX and pick the Overwrite policy that matches your trust level for the incoming pairs.

[Screenshot: The Visual Editor showing a translation segment with a Translation Memory suggestion panel displaying a 92% fuzzy match with highlighted differences]

—