How to Translate a Bates-Numbered PDF Without Breaking Legal References
Translate a Bates-numbered PDF with a frozen source inventory, page-level mapping, controlled identifiers, and final-file reference checks.
To translate a Bates-numbered PDF safely, treat every Bates identifier as a reference key, not as prose. Freeze the original production, record one row per source page, translate from a working copy, and prove how every target page maps back to the source identifier.
Do not invent a target numbering convention during translation. A contract, production protocol, court order, client instruction, or review procedure may control whether the translation keeps the original identifier, adds a suffix, or uses a separate range. The responsible legal or records team must approve that rule before work begins.
Use this sequence:
- Preserve the exact source files and their hashes.
- Inventory filenames, file order, PDF page indexes, and visible Bates identifiers.
- Resolve missing, duplicate, illegible, or conflicting identifiers before translation.
- Approve a target identifier rule, including what happens when one source page becomes several target pages.
- Translate the language content without treating the identifier as translatable text.
- Build a source-to-target page map.
- Test every citation and cross-reference against the final files.
Download the Bates translation page-map template. It includes an illustrative split-page pattern, not a universal naming rule.
What a Bates Number Does
A Bates number is a unique page identifier used across a set of documents. It is usually rendered in a header or footer and can combine a numeric sequence with a prefix or suffix.
Adobe's current Acrobat instructions describe a batch process in which users choose the file order, starting number, number of digits, prefix, suffix, output folder, and filename behavior. That file order matters: changing it can change which page receives which identifier.
Keep these concepts separate:
| Identifier | Example | What it identifies |
|---|---|---|
| Bates identifier | ACME000247 | One page in a controlled document set |
| PDF page index | Page 18 in the file | The page object's position in that PDF |
| Printed page number | 12 printed in a footer | A number in the document's own content |
| PDF page label | A-12 shown by a viewer | A logical label mapped to a PDF page |
| Filename | production-003.pdf | The container holding one or more pages |
These values can disagree without any of them being corrupt. A PDF can have six unnumbered cover and index pages, printed page 1 at PDF page 7, and Bates identifier ACME000247 on that same page. The Bates identifier remains the external reference key.
Our PDF page-number repair guide owns printed pagination and PDF page labels. This guide owns the separate source-to-target Bates relationship.
Freeze the Source Production Before Translation
Do not start from a folder whose contents can still change. Create a read-only source set and record:
- production or matter name;
- source filename and SHA-256;
- file order;
- page count;
- first and last Bates identifier;
- prefix, suffix, and digit width;
- confidentiality, redaction, or status markings;
- whether the PDF is encrypted, signed, scanned, or form-based; and
- the instruction that authorizes translation and defines the deliverable.
Adobe notes that protected or encrypted PDFs and some form types are not supported by its Bates-numbering operation. That is a useful preflight distinction: an edit restriction, form behavior, signature, or encryption issue should be resolved by the authorized file owner rather than hidden by printing the document into a new PDF.
Hashing matters because filenames are weak evidence. Two files named production-001.pdf can contain different revisions. A page map tied to a checksum lets a reviewer determine which bytes were actually translated.
Audit the Existing Sequence
Build one inventory row per source page before modifying the PDF. Capture at least:
source filename
source file hash
PDF page index
visible Bates identifier
identifier location
printed page number or page label, when relevant
page status: present, blank, redacted, duplicate, illegible, or exception
Then test the sequence mechanically and visually.
Look for gaps and duplicates
A sequence such as ACME000101, ACME000102, ACME000104 has a gap. That does not prove a page is missing: the omitted identifier may belong to another file, may have been intentionally withheld, or may reflect an earlier production decision. Record the gap and resolve it against the production index instead of silently renumbering the pages.
A duplicate identifier is equally important. Determine whether the duplicate is a source defect, a replacement page, or a legitimate convention. Never fix it only in the translation; that would leave the source and target referring systems inconsistent.
Check what is actually visible
Inspect the first page, last page, every file boundary, and a sample from the middle. Also inspect pages with redactions, rotated content, very dark scans, or dense footers. OCR or extraction can miss an identifier even when a person can see it, while an identifier added as ordinary PDF text can be extracted in the wrong reading order.
Use visual inspection to establish presence and legibility. Use text extraction as a second signal, not the sole authority.
Approve the Target Identifier Rule
There is no universal translation suffix or pagination rule. Choose one controlled policy and record who approved it.
Common patterns include:
| Pattern | Illustration | Main tradeoff |
|---|---|---|
| Keep the source identifier visibly unchanged | ACME000247 | Simple reference continuity, but source and translation need another visible distinction |
| Keep the source identifier and add an agreed translation suffix | ACME000247-T | Direct relationship, but the suffix format must be governed |
| Assign a separate target range and maintain a map | FR000811 → source ACME000247 | Clear target set, but every citation depends on the map |
| Use source identifier plus page-part suffix after expansion | ACME000247-T.01, .02 | Supports reflow, but the part convention must be defined in advance |
A real case protocol can be more specific. For example, a published translation stipulation in a U.S. federal case required translated pages to retain the source Bates number, add a translation marker, and use decimal-style additions when one source page required multiple translated pages. That is evidence that page expansion needs an explicit rule; it is not a default for other matters.
The controlling instruction may also require a Translation label, retention of confidentiality designations, a certification page, or a particular filename. Copy those requirements exactly into the release checklist. Do not infer them from another case or client.
Do Not Translate the Identifier as Language
A prefix can look like an abbreviation or word, but its role may be structural. PLAINTIFF000247, FACTURE000247, or JP-EXH-000247 should not be localized merely because the surrounding document is being translated.
Lock these fields unless the approved policy changes them:
- prefix and suffix;
- numeric value and zero padding;
- capitalization and separators;
- identifier placement;
- associated filename or load-file key; and
- replacement or translation designators.
Keep identifier strings out of normal machine-translation segments when possible. If they must pass through the same extraction pipeline, place them on a protected-term list and compare the extracted source/target identifier sets afterward.
The same rule applies to citations inside the text. A sentence such as “See ACME000247” contains translatable prose and a nonlinguistic reference key. Translate the sentence; preserve or deliberately remap the key.
Map Every Target Page Back to the Source
The target can preserve page count, shrink, or expand. Do not assume one source page always equals one target page.
The downloadable CSV uses one row per target page relationship. The core columns are:
| Column | Purpose |
|---|---|
source_file_sha256 | Pins the map to the exact source file |
source_pdf_page_index | Locates the page inside that file |
source_bates_id | Preserves the external reference key |
target_file_sha256 | Pins review to the final translated bytes |
target_pdf_page_index | Locates the translated page |
target_bates_id | Records the approved visible target identifier |
translation_page_part | Explains a one-to-many relationship |
identifier_decision | States the policy applied to that row |
content_check | Records whether the page's translated content passed review |
reference_check | Records whether citations and references resolve correctly |
Never compress a one-to-many relationship into a vague note. If source page ACME000247 becomes two target pages, create two target rows that both point back to ACME000247 and carry the approved part identifiers.
Likewise, if two short source pages are deliberately composed onto one target page, record both source relationships. The map should explain the target, not force reality into a one-to-one spreadsheet shape.
Translate from a Working Copy
Keep the source production immutable. Create a working copy for translation and decide how the visible identifier will survive the workflow.
Before using any translator, test a representative file containing:
- a normal text page;
- a rotated or landscape page;
- a scanned page requiring OCR;
- a redacted page;
- a page with a confidentiality mark;
- a page where the Bates identifier sits close to body text; and
- a page with an internal citation to another Bates identifier.
Check whether the output preserves, removes, duplicates, obscures, or mistakenly translates the identifier. Also check whether target-language expansion pushes body text over it.
If the identifier is part of the page image, OCR may recognize it as body text. If it is a header/footer object, reconstruction can omit it. If it is added again after translation, an undetected surviving copy can create two identifiers on one page. The only safe assumption is that the exact output must be inspected.
For image-only productions, use the scanned PDF translation workflow to establish OCR quality first. For general layout defects, use the format-preserving PDF workflow.
Verify References, Not Just Page Counts
A target PDF can have the expected page count and still break the purpose of Bates numbering. Test references at three levels.
1. Identifier integrity
- Every source identifier appears in the source inventory.
- Every target page has the identifier required by the approved policy.
- No target identifier is duplicated unintentionally.
- Prefix, suffix, digit width, separators, and case match the policy.
- The identifier is legible and does not obscure document content.
2. Mapping integrity
- Every target page maps to at least one source page.
- Every source page has an expected target disposition.
- Split and combined pages have explicit relationships.
- Deleted, withheld, blank, or untranslated pages have recorded decisions.
- File boundaries and page order agree with the approved target set.
3. Citation integrity
Search the final translation for every Bates-style reference. For each citation, open the referenced source or target location and confirm that it supports the surrounding statement. Do this for:
- references within the translated documents;
- an index or chronology;
- review notes and issue logs;
- declarations, reports, or correspondence delivered with the translation; and
- any export produced from a review platform.
If the target uses a new range, reviewers need the map at the point of use. A correct map stored in a private folder does not help someone reading a translated report that cites only the new identifier.
Run Failure-Injection Checks
A good QA process should catch deliberate defects before it is trusted on real material. Copy a small synthetic batch and introduce one error at a time:
- remove a target page;
- swap two files in the batch order;
- duplicate an identifier;
- change one prefix or digit;
- let one source page expand without part identifiers;
- leave one internal citation pointing to the wrong page; and
- cover part of an identifier with translated text.
Run the inventory and reference checks. If a defect is not detected, improve the control before using it on a larger production.
The CSV is intentionally a template rather than a claim that these checks ran on BookTranslator output. Fill it with the actual hashes, decisions, reviewers, and results for your files.
Where BookTranslator Fits
BookTranslator PDF Translator can translate the language content of a supported PDF when you have permission to process it. It is not a Bates-management system, litigation platform, legal certification service, or source-of-truth production database.
Before uploading, preserve the source production and its page inventory. Start with one representative, authorized file rather than the whole production, then determine how the output handles the visible identifier, redactions, stamps, page geometry, and internal references. After downloading, treat the translated PDF as a candidate output: hash it, populate the page map, and run identifier and citation checks on the exact final bytes.
Do not claim that a successful upload proves that Bates numbering, redactions, signatures, forms, annotations, bookmarks, or evidentiary relationships were preserved. Each is a separate release surface. The PDF translation QA checklist covers the broader final-file review, while the PDF bookmark guide covers navigation destinations.
If the workflow requires certified translations, declarations, load files, privilege handling, or court-specific formatting, assign those requirements to the qualified people and systems responsible for them.
Release Gate
Do not release the translated set until:
- the source file list, order, page counts, and SHA-256 values are frozen;
- every source page has a recorded Bates identifier or documented exception;
- gaps and duplicates are resolved against the source production record;
- the target identifier and page-expansion policy has an identified approver;
- every target page maps back to its source page or pages;
- target filenames and identifiers follow the approved convention;
- identifiers are legible, unique where required, and do not obscure content;
- redaction and confidentiality markings have separate pass/fail checks;
- internal and external Bates citations resolve to the intended content;
- the exact final PDFs, not an intermediate export, have recorded hashes;
- the broader content and layout QA has passed; and
- unresolved exceptions name an owner and prevent release when they affect reference integrity.
The governing principle is simple: translate the language, preserve the reference system, and prove the relationship page by page.
Related Posts
