LoginSign Up
← Back to all sparks
L

Lokalise

DEVOPS
Velocity5.0

Lokalise is instrumenting AI translation quality so teams can see how much human correction it costs.

localizationtranslation-qualityai-mtanalyticstranslation-memorytooling
◆Current state
Lokalise is concentrating on measuring and improving translation quality in AI/MT-heavy workflows: a Translation Quality Analytics beta tracking post-edit rate and edit distance, richer per-contributor review metrics, smarter Translation Memory that now captures reviewer-approved AI/MT output, and a browser-based Glossary Guard for cleaning glossary files. Performance and tooling work (faster snapshots, a rewritten Go file-exchange library) rounds it out.
◆Where it's heading
The direction is quality measurement as the control layer over machine translation: give localization managers hard numbers on how much post-editing AI output requires, and feed validated output back into TM to compound. Lokalise is positioning around trust in MT output rather than just generating more of it.
◆Prediction
Expect the Translation Quality analytics to graduate from beta and tie more directly into TM and workflow routing, surfacing where AI/MT is reliable enough to auto-approve versus where human review pays off.

◆Recent moves

  1. 9d ago

    Glossary Guard is now available as a web app

    Glossary Guard, the tool that checks and fixes glossary CSVs before upload, is now a browser-based web app in pre-release, no install required. Broadens access to a previously terminal-only utility.

  2. 25d ago

    Human-reviewed AI/MT translations now saved to Translation Memory

    Any AI/MT translation a reviewer marks Reviewed, even without edits, is now saved to Translation Memory under their name, not just manually edited ones. Tightens how validated AI output compounds into reusable TM.

  3. 1mo ago

    Translation Quality Analytics is now available in Open Beta

    Translation Quality Analytics opens in beta with post-edit rate, average edit distance, and review-effort distribution across languages and workflows. The measurement layer at the center of the quality-instrumentation push.

  4. 2mo ago

    Richer review metrics in Task Analytics

    Task Analytics adds seven per-contributor review metrics including acceptance rate, edit buckets, review turnaround, and average AI score. Gives managers granular insight into review effort and AI output quality.

  5. 2mo ago

    Richer review metrics in Task Analytics

    A near-duplicate of the same Task Analytics review-metrics release, republished with a slightly different timestamp. Same content, no additional change; counted once in the trajectory.

    View source ↗
  6. 2mo ago

    Filter Task Analytics by creation or completion date

    The Tasks Time Analytics dashboard can now filter by creation or completion date instead of defaulting to creation only, available on all plans. A small but real reporting-flexibility gain consistent with the analytics focus.