LoginSign Up
← Back to all sparks
L

Lokalise

DEVOPS
Velocity5.0

Lokalise is instrumenting the human review layer around AI translation — quality, not just throughput.

localizationtranslation-memoryai-translationquality-analyticsreview-workflowsglossary
◆Current state
Lokalise is building out the review-and-quality side of AI/MT-driven localization. Recent releases automate how translation-memory matches flow through workflows, capture human-approved AI/MT into TM, and add analytics that measure post-editing effort and translation quality — plus a self-serve Glossary Guard web app and much faster project snapshots.
◆Where it's heading
As machine and AI translation take over raw volume, Lokalise is recasting the human job as review and QA and instrumenting exactly that: TM automation to cut redundant review, and quality analytics (post-edit rate, edit distance) to show where AI output can and can't be trusted. The direction is a measurable, leaner AI-assisted localization pipeline.
◆Prediction
Expect Translation Quality Analytics to move from open beta toward GA, with tighter loops between quality signals and workflow automation — for example auto-routing low-confidence segments to human review.

◆Recent moves

  1. 3d ago

    More control over TM matches and review scope in Workflows

    Adds two Workflow controls — auto-setting the status of TM-matched translations and excluding already-reviewed segments from review tasks — so TM matches can run fully hands-off. Directly serves the goal of cutting redundant review work.

  2. 21d ago

    Glossary Guard is now available as a web app

    Ships Glossary Guard as a no-install browser app, opening a previously CLI-only cleanup tool to non-developers on localization teams. Reduces upload errors, though still pre-release.

  3. 1mo ago

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

    Now saves any AI/MT translation a reviewer marks Reviewed into Translation Memory, not just edited ones — capturing human sign-off as reusable, traceable TM. Grows TM leverage as AI handles more of the first pass.

  4. 1mo ago

    Translation Quality Analytics is now available in Open Beta

    ⚡ SPARK

    Opens a Translation Quality section in Analytics measuring post-edit rate, edit distance, and review effort across languages and workflows — the first real view into whether AI/MT output is good, not just how much shipped. Anchors the shift from throughput metrics to quality metrics.

  5. 2mo ago

    Richer review metrics in Task Analytics

    Adds seven per-contributor review-quality columns — acceptance rate, edit buckets, average AI score, review turnaround — to Task Analytics. The granular building blocks feeding the broader quality-analytics push.

  6. 2mo ago

    Richer review metrics in Task Analytics

    A duplicate changelog capture of the same 'Richer review metrics in Task Analytics' release as the prior entry, with no additional change. A same-release feed duplicate.

    View source ↗