Translation & qualityBLoc PDF QC11 min read

How to Turn Translated PDF Reviews into Clear DTP Corrections

Compare source and translated PDFs side by side, triage layout and text findings, and export confirmed corrections as annotated PDFs and Word or Excel logs.

Executive summary · 11 min read
  • Set layout direction (LTR/RTL) and a project brief before comparing, so approved differences stop becoming noise.
  • Findings are candidates with evidence: reviewers confirm, resolve, or ask - nothing is auto-approved.
  • Confirmed Open/Ask comments survive rescans and revised targets; check new locations before exporting.
  • Export only confirmed comments to annotated PDF plus Word and Excel logs for the DTP team.
Local AI & Relocation Verification: Optional local AI review in BLoc PDF QC never silently falls back to a cloud provider. When loading a revised target PDF, comments relocate automatically only where evidence supports a unique match; visually verify scanned pages, embedded image text, and any ambiguous relocations before exporting your handoff package.

A translated PDF lands in review, someone screenshots a few suspicious lines, and the screenshots go to the DTP operator with a sentence like "fix these". Two days later the same findings come back, because the screenshots never said what was wrong, where it was wrong, or whether the fix was agreed. The gap is not effort; it is structure. BLoc PDF QC exists to put structure into that handoff: a side-by-side comparison of source and target, review candidates you inspect and confirm, and an export of confirmed comments that a DTP operator can execute without guessing. This guide covers the review workflow from baseline setup to the annotated PDF and the Word or Excel correction log.

Why raw diffs fail the DTP handoff

Comparing two PDFs by eye produces three kinds of noise that make handoffs unreliable. First, raw differences are not defects: a justified paragraph that ends one line earlier in Arabic is typography working, not an error. Second, eyeballing has no measured basis — "the line looks tight" is not an instruction an operator can act on. Third, unstructured findings have no state: nobody can tell a confirmed defect from an open question, so both arrive on the operator's desk with the same weight and the important ones drown.

A production review inverts those failures. The source document acts as the baseline, differences become review candidates with evidence attached, and a reviewer moves each candidate through explicit states until only confirmed corrections remain. The export then contains exactly what has been decided, which is why the DTP team can trust it.

Set the baseline: layout direction and project brief

Before comparing anything, tell the tool how to read the documents. BLoc PDF QC compares source and target side by side with an explicit layout direction — LTR or RTL — set for the project. Getting direction right is not cosmetic: in an RTL target, reading order, spread order, and the meaningful side of a margin all flip, and a comparison engine that assumes the wrong direction will generate nonsense candidates on every page.

The project brief serves the same purpose for humans. It records what this document is, which decisions were intentional, and what the reviewers should prioritize. When a candidate contradicts the brief, the brief wins the argument — or the brief gets updated through a decision, not silently. Reviews that skip this step tend to rediscover the same approved differences on every revision.

Layout findings and measured typographic clues

BLoc PDF QC inspects the target against the source and surfaces candidates across concrete categories. Knowing what each category measures tells you where the underlying production fault lives:

Finding categoryWhat is surfacedTypical production cause
OverlapCandidates where content collidesTarget text expands past its frame into a neighboring element
Page edgesContent at or beyond page boundariesReflow pushes text or tables past the trim or safety margin
ImagesImage-related candidatesImages missing, moved, or swapped during re-layout
Page numberingNumbering differencesReflow changed pagination; static numbers no longer match
Printed spreadsSpread composition findingsSpread order or composition broken for the target direction
PunctuationPunctuation differencesWrong-direction marks or substituted punctuation in the target
Line spacingMeasured spacing cluesFont substitution changed leading without review
Character spacingMeasured spacing cluesTracking applied to fit text, distorting the type
Font sizeMeasured size differencesText shrunk to fit instead of being restructured
HyphenationMeasured hyphenation cluesHyphenation rules applied across languages inconsistently
HyperlinksBoundary and underlining cluesLink text rewrapped, breaking boundary or underline integrity
Source/target differencesGeneral text differencesAnything from real omissions to approved localization choices

Alongside layout findings, spelling and numeric consistency candidates appear with the evidence for each, so a reviewer can decide rather than decode. The measured clues matter because they convert taste arguments into checkable facts: a spacing finding either survives measurement or it does not.

Reviewer triage: candidates, confirmation, and Open/Ask status

Automated checks and the optional local AI review produce candidates — never verdicts. The optional AI review runs locally with page evidence and source/target matching, saves checkpoints, and can be cancelled and resumed, which matters on long documents. But whether a finding came from a deterministic check or an AI pass, it becomes a correction only when a reviewer inspects the evidence and confirms it.

Triage is a filtering job. Draw a precise comment box on the exact element, then use the filters — page, category, severity, status — to work through findings in batches of similar kind. Edit comments that misdescribe the problem, confirm the ones that are real, resolve the ones already handled, and undo deletions when a comment was removed in error. Each comment lands in one of two export-facing states: Open, a confirmed correction for DTP to apply, or Ask, a question that needs an answer before anyone touches the file. Keeping those two states separate is the difference between a correction list and a conversation.

Production note: Unconfirmed findings that require manual review stay out of exports by design. If an item is not ready to be a correction, it should not be in the DTP package — leaving it unconfirmed is the mechanism, not a limitation to work around.

Keeping comments alive across rescans and revised targets

Real reviews are loops, not passes. The DTP team applies fixes, a revised target arrives, and the review starts again. BLoc PDF QC is built for that loop: accepted and manual comments survive rescans and revised targets. When a revised target is supplied, the tool checks the comments' new locations — and a reviewer verifies those locations before export, because reflow can move the surrounding content even when the correction was applied correctly.

The same persistence applies to supported Word documents reviewed through the tool: comments can be revisited when a revised document is supplied. In practice, treat every revised target as a delta review — confirm that confirmed comments moved to where you expect, look at neighboring content affected by reflow, and only then export the updated package. This is also where the audit trail pays off: a comment that has survived three revisions with its history intact is far more credible than a fresh sticky note.

Exports DTP can act on: annotated PDF, Word, Excel

The handoff artifact set has two halves. The annotated target PDF carries numbered comment boxes at precise locations so the operator can find each item visually; it is produced by a background export job with progress and retry support, so long documents do not lock the reviewer's session. The Word and Excel comment lists carry the same confirmed Open/Ask comments as structured records for assignment, tracking, and status columns in the project tracker.

Before sending, verify the package the way you would verify a build: every marker on the annotated PDF corresponds to one record in the list, statuses are current, and no stale comments from a previous revision are still marked open. Originals are preserved and reviews are saved locally, so the export is a view of the review state, not a mutation of your files.

Pre-delivery engineering checklist

  1. Confirm the project's layout direction (LTR or RTL) matches the target, and the project brief reflects current decisions.
  2. Work candidates by category with filters, and attach precise comment boxes rather than page-level notes.
  3. Classify every kept comment as Open (confirmed correction) or Ask (open question) — never mix the two in the export.
  4. After any rescan or revised target, check each surviving comment's new location before exporting.
  5. Review spelling and numeric consistency evidence and record a decision for each candidate.
  6. Generate the annotated PDF via the background export and confirm it completes without retry warnings.
  7. Cross-check Word/Excel records against PDF markers one-to-one, and confirm statuses are current.
  8. Archive the review state locally with the delivery, so the next revision starts from history.

Technical FAQ

Does the tool decide which corrections to apply to the PDF?

No. Automated checks and the optional local AI review identify review candidates with evidence. You inspect each candidate and decide which comments to confirm. Unconfirmed findings that require manual review stay out of exports, so the DTP team receives only reviewer-confirmed items.

What happens to my comments when a revised target arrives?

Accepted and manual comments are kept through rescans and revised targets. The tool checks the comments' new locations in the revised document, and you verify those locations before exporting. Treat the revised pass as a delta review: confirm surviving comments, look at reflow-affected neighbors, and update statuses.

Why is one of my findings missing from the Excel export?

Exports include confirmed Open/Ask comments. A finding that was never confirmed, or that is still flagged as requiring manual review, is intentionally excluded. Confirm the comment — after inspecting its evidence — and it will appear in the next export.

Can I pause a long AI review and resume it later?

Yes. The optional local AI review saves checkpoints and provides cancel and resume controls, so a long document can be reviewed across sessions without losing progress. Evidence and source/target matching are retained, and you still confirm or discard each AI-surfaced candidate yourself.

For the surrounding pipeline, see the BLoc PDF QC tool page and the hands-on interactive walkthrough. Where translated content still needs to get into design files first, the companion guide on file preparation and XLIFF exchange covers the stage before typesetting, and batch translation with terminology and templates covers the stage before that.

BLoc PDF QC · Document quality

Ready to apply this workflow in production?

Compare source and translated documents, review quality findings, and export confirmed corrections for DTP. Pricing: $20 / month or $200 / year.

← All 14 localization engineering guides