Translate | ISO 19005-1:2005 PDF/A-1 Archival Documents, Embedded Fonts, Preservation Claims and Multilingual Records

Anyone who needs to translate PDF/A-1 archival documents under ISO 19005-1:2005 must solve two problems at once: preserve the meaning of the source record across languages and create a dependable long-term electronic file. A clerk may be translating an archived certificate; a university may need a bilingual historical record; a court-support team may need a controlled translated exhibit; a museum may publish multilingual inventories. Their search intent is practical: how to translate a PDF/A-1 file, retain document identifiers and signatures, embed fonts, validate the resulting PDF/A and distinguish the original record from the new translation.

ISO 19005-1:2005 is titled Document management — Electronic document file format for long-term preservation — Part 1: Use of PDF 1.4 (PDF/A-1). The official ISO catalogue records this edition as published and confirmed in 2026. The public description concerns using PDF 1.4 for preservation of documents combining text, graphics and images. PDF/A-1 has A and B conformance levels: the basic B level focuses on reliable visual reproduction, while A includes additional requirements supporting structure and text interpretation. Neither level guarantees that a translation is correct, an OCR transcript is accurate, a record has legal effect in another jurisdiction or that a historical digital signature survives editing.

Quick answer: keep the source archive immutable, prepare the translation as a separately identified derivative record, verify every name, date, seal, table, reference and qualification, then validate the actual translated PDF/A-1 file using a suitable conformance checker. Inspect visual rendering, embedded fonts, searchable text, character mapping, metadata and document relationships by hand as well. Preserve a chain between original and translation; do not present a language edition as though it were the untouched source. This article provides original worked cases, a quality-control workflow, practical terminology and a four-week learning plan for multilingual archival work.

Quick answer for archival teams

• Keep an immutable source record and give the translation its own controlled identifier.
• Check every name, date, number, table, seal, source qualification and uncertain OCR fragment.
• Distinguish PDF/A-1a from PDF/A-1b and validate the final file.
• Inspect font embedding, rendered glyphs, text extraction, metadata and source relationships.
• Do not confuse file-format validation with linguistic, legal or historical certification.

Contents

1. The core decision: original, rendition or translation
2. Why PDF/A-1 matters for multilingual records
3. PDF/A-1a and PDF/A-1b are different claims
4. Archive-first translation workflow
5. Fonts, glyphs and character identity
6. Scanned pages and OCR are not translated evidence
7. Metadata is a record of provenance, not decoration
8. Signatures, seals and authentication
9. Colours, layers, transparency and interactive features
10. Tables, units and legal qualifiers
11. Language identification and accessible reading
12. Validation is necessary and insufficient
13. Change management without overwriting history
14. One specialist answer, not another translation hub
15. Controlled terminology
16. 32 practical examples
17. Release review, FAQ, and four-week learning plan

The core decision: original, rendition or translation

A reader cannot evaluate a translated archival PDF until they know which role the file plays. An original may be a source record kept under an institution’s retention rules. A rendition might reproduce the same content in a different technical format. A translation changes the language of the record; by its nature, it is a new interpretive expression, even when closely faithful. Treat these as distinct assets with distinct identifiers, creation dates, responsible people and provenance. The translator should not overwrite the original or imply the derivative is an authenticated original. In a legal or administrative context, the institution may require a signed translation statement, but that local requirement is separate from PDF/A compliance. Keep archive and language-approval decisions in their proper lanes.

Why PDF/A-1 matters for multilingual records

Long-term readability depends on more than a file extension. A PDF that displays correctly on one workstation may rely on a font, colour setting or interactive feature unavailable decades later. PDF/A-1 is designed to limit some such dependencies. For translators this matters because Vietnamese accents, Arabic shaping, Cyrillic, Chinese characters, mathematical symbols and specialised marks can become invisible or corrupted when the wrong font is used. A visually correct printout can also hide broken text extraction. The practical sequence is to preserve source meaning, select embeddable fonts with required glyph coverage, create a stable layout, and validate both the standard’s technical properties and actual reader experience. The translator should not declare longevity guaranteed.

PDF/A-1a and PDF/A-1b are different claims

The Library of Congress identifies two PDF/A-1 conformance levels. PDF/A-1b aims at preserving visual appearance. PDF/A-1a adds requirements related to document structure and text properties that facilitate reliable interpretation. A file that passes level B does not thereby satisfy level A. A file that passes level A still may have incorrect alternative text, illogical reading order or mistranslated content, which require human review. When translating a validation report, preserve the exact designation, not merely ‘PDF/A compliant’. Readers need to know which checks were performed and with what tool, profile and version. Do not infer accessibility certification from a PDF/A label.

Archive-first translation workflow

Keep the source bitstream and its archive identifier under the repository’s ordinary integrity controls. Record where it came from, when received and what the source file actually states. Extract the content carefully, including page headers, footnotes, exhibits and marginal notes. For scans, distinguish human transcription from machine OCR and record uncertain glyphs. Translate using a controlled source text, not only a copied paragraph that may have lost tables or superscripts. Run language review and domain review. Create a new PDF from an approved target master rather than editing the archived original in place. Register both files with a documented relationship, their language codes and release status. Finally validate and inspect the derivative.

Fonts, glyphs and character identity

Embedding a typeface helps preserve appearance, but the translator still must check every glyph. An Arabic numeral, minus sign, Greek letter or accented name may map incorrectly in text extraction. A font can show a symbol visually while a copy-and-paste operation returns the wrong character; that matters for search and later migration. Build representative test pages covering the exact target script, diacritics, ligatures, mathematical characters and punctuation. Confirm embedding rights and successful subset inclusion where applicable. Avoid substituting decorative glyphs for semantic characters if readers need to search the text. Keep a record of font and export settings so the derivative is reproducible rather than dependent on one desktop installation.

Scanned pages and OCR are not translated evidence

Optical character recognition is a fallible reading of an image. It may turn 8 into B, confuse a date separator, omit a handwritten negation or merge table columns. A translator must compare extracted text to the visible source and flag ambiguous characters. If the document is handwritten, smudged or damaged, mark an illegible field rather than inventing the missing word. An OCR layer may aid search but not prove the text was accurately captured. Before translating a certificate or archival register, assign each uncertain fragment a review status and require human inspection. Keep the image of the original page accessible through controlled archival reference. The translated PDF should not launder uncertain OCR into unquestioned certainty.

Metadata is a record of provenance, not decoration

PDF/A-1 uses XMP metadata to carry standardised document information. For a translation, metadata should record the target-language document title, author or producing organisation where appropriate, source relationship, version and meaningful dates under local repository rules. Do not overwrite the date an historical deed was signed with the date the translation was created; both can exist in different fields. When the original contains identifying or confidential data, follow the applicable governance and access rules instead of indiscriminately copying metadata. A PDF/A declaration field is not a declaration that content was examined or certified. Validate XMP consistency with the file’s claimed conformance level and document information fields.

Signatures, seals and authentication

A translated signature label or a reproduced seal is not an actual signature by the person named. Distinguish ‘signature visible in source image’, ‘signed electronically’, ‘verified signature’, and ‘certified translation’. Editing or converting a signed PDF may invalidate its cryptographic signature. The correct workflow often preserves the signed original untouched and attaches the translated derivative as a separate record, with an independent statement of translation if legally required. Do not insert an imitation signature into a target document to make it look official. Note illegible seals, broken certificate chains or missing validation evidence openly. PDF/A validation assesses file-format properties, not signatory identity, legal admissibility or provenance authenticity.

Colours, layers, transparency and interactive features

PDF/A-1 forbids certain features that could compromise dependable future reproduction, including transparency, encryption and JavaScript. A translated document designer may accidentally add a soft-shadow logo, dynamic widget, unembedded font or script-based form field. Validate the exact exported file, not just its design source. A flattening operation may change appearance, obscure small footnotes or harm text extraction, so compare renderings page by page. A scanned colour seal and a translated plain-text explanation serve different functions. Do not alter source evidence to satisfy layout preferences. When genuinely incompatible features are required, revisit the chosen preservation profile and repository requirements rather than falsely advertising PDF/A-1.

Tables, units and legal qualifiers

Multilingual archive translation is particularly vulnerable in tabular records. A ledger row may represent a debit, credit, unit, tax reference or page total; its meaning can be lost if columns shift or a minus sign disappears. Preserve the source order and identifier fields, adding a glossary or explanatory note outside the evidentiary text if authorised. Numbers that look like dates may follow different local conventions. Put explicit months in translator notes where ambiguity would matter, but do not silently reinterpret an unclear source date. A legend such as ‘copy’, ‘draft’, ‘issued’, ‘void’, ‘certified’ or ‘not for official use’ must survive the target version with the same force.

Language identification and accessible reading

Metadata can record the language of a document, and structured PDFs can assist assistive technologies, but accessibility is not one yes/no property. A bilingual PDF may have paragraphs in two languages, tables mixing scripts and quoted original phrases. Test reading order, text selection, headings, language switching and copied text with the actual output tools and assistive workflows in use. If the conformance level does not guarantee a meaningful structure, don’t claim otherwise. Also avoid hiding a critical translator’s note as an inaccessible image. Readers using magnification, screen readers or text extraction should be able to distinguish the original text from explanatory translation commentary.

Validation is necessary and insufficient

Run a capable PDF/A validator against the final bytes and save the exact report. Investigate failures such as missing fonts, broken XMP claims, unsupported features or invalid colour settings. But a technical pass only addresses properties checked by its validation profile; it cannot detect an incorrect translated surname or swapped debit amount. Perform a second audit: inspect every page visually, verify extraction of sample terms in all target scripts, compare tables, and cross-check translated notices against source. Add provenance and evidence checks to the release worksheet. Do not treat a green software badge as certified translation or archival policy compliance.

Change management without overwriting history

An archive must support historical reconstruction. When a translator corrects an earlier version, create an authorised new derivative with a version statement, not a silent replacement of what a reader previously received. Keep the original, earlier translation, correction reason and release dates as the repository requires. A typo correction to a heading and a revised interpretation of a financial amount are not equivalent changes; the second may require domain approval and user notification. If a future migration changes PDF/A flavour, register the transformation as a new technical rendition linked to the same intellectual content. Provenance is a chain of facts about objects and processes; it is not an excuse to revise historical evidence.

One specialist answer, not another translation hub

The search intent of this article is ‘How can I translate a PDF/A-1 archival record, keep meaning and prove the file meets the selected archival format?’ It belongs alongside narrower guides about filenames, metadata and legal records, not in competition with the translation master. Readers should immediately find the file-format distinction, a translation workflow, practical examples and validation limits. A finished work package includes the controlled source, translated text master, final PDF/A derivative, validation results, linguistic and domain approvals, metadata record and link to the authoritative archive. Use precise indexing language such as PDF/A-1a, PDF/A-1b, embedded fonts, OCR verification and XMP so readers reach the right information.

Twenty source concepts to translate with care

These are plain-language source meanings, not universal translations into any specific language. Select target expressions with qualified reviewers, noting the differences between a format property, a document-management fact and a linguistic conclusion.

PDF/A-1

PDF archival profile based on PDF 1.4 under ISO 19005-1; it is a format designation.

PDF/A-1a

The more demanding PDF/A-1 conformance level, including additional structure and text-representation requirements.

PDF/A-1b

A PDF/A-1 conformance level focused principally on dependable visual reproduction.

Source record

The original preserved record on which any derivative rendition or translation is based.

Derivative record

A new work product based on a source, including a separately identified translation.

XMP

Extensible Metadata Platform metadata embedded in the PDF for described properties and format declarations.

Embedded font

A font program or appropriate subset included within the file for reliable text rendering.

Glyph

The visual shape of a character; the shape may be correct even when the encoded character is wrong.

ToUnicode mapping

A PDF mechanism relating character codes to Unicode for extracting and interpreting text.

OCR

Optical character recognition that estimates text from page images; must be verified.

Text layer

Machine-readable text associated with a page, which may differ from the visible image.

Reading order

The meaningful sequence in which text and structural elements should be consumed.

Conformance validator

Software that checks whether a specific PDF satisfies a particular PDF/A validation profile.

Digital signature

A cryptographic signing mechanism whose integrity can be affected by altering the file.

Output intent

Colour-related metadata helping define intended rendering conditions.

Translation provenance

Record of which source, revision and human process produced a target-language derivative.

Redaction

Controlled removal of information, distinct from merely hiding text visually.

Preservation copy

A managed archival object with defined provenance, fixity and retention responsibilities.

Fixity

Evidence that a digital object has not changed unexpectedly, often evaluated with a checksum.

Conformance claim

Statement that a given final file meets a specified version and level of PDF/A.

32 original worked archival translation cases

Each case is an educational scenario. The text is invented to reveal common risks; it is not part of the ISO normative language, and a translation of a real official record must follow the competent authority’s documentary rules.

Case 1: Not a replacement for the original

Illustrative source: “Certified translation attached; original retained.” Target-language meaning to retain: “A translated derivative is attached while the source record remains preserved unchanged.”

The status of the translation and the fate of the original are two separate facts. Translating ‘attached’ as ‘replaced’ could imply a custody action that never happened. Check the archive identifier and custody register. Release check: Do not present the derivative as the authenticated source.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 2: Wrong PDF/A level

Illustrative source: “Validated as PDF/A-1b.” Target-language meaning to retain: “The file passed checks under the PDF/A-1b profile.”

A PDF/A-1b pass does not establish level A conformance or accessibility properties. Keep the letter suffix as part of the format designation. Compare validator profile and exact report. Release check: Do not change 1b to 1a for marketing.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 3: Validation still pending

Illustrative source: “PDF/A compliance check pending.” Target-language meaning to retain: “The file has not yet completed the planned conformance validation.”

An export preset can claim PDF/A but the output can still contain errors. Only a final-file validation result supports the stated status. Review validation job log and output bytes. Release check: Do not say ‘archival compliant’ before testing.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 4: OCR confidence

Illustrative source: “OCR reads ’18’ or ‘1B’; visual confirmation required.” Target-language meaning to retain: “The extracted characters are ambiguous and require checking against the scan.”

Numbers and letters may look almost identical in damaged historical records. Choosing one without review can change an account or identity. Inspect source image at suitable resolution and note uncertainty. Release check: Do not silently pick the apparently sensible value.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 5: Name diacritics

Illustrative source: “Surname appears as García; OCR renders Garcia.” Target-language meaning to retain: “The source shows the accented name García, while OCR omitted the accent.”

A proper name is not a decorative field; missing marks can affect matching against other records. Preserve source orthography with a note when target systems constrain characters. Verify image, identity source and Unicode output. Release check: Do not normalise names solely for convenience.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 6: Source date versus translator date

Illustrative source: “Instrument dated 1908; translation completed 2026.” Target-language meaning to retain: “The historical record carries its 1908 date; the translation was produced in 2026.”

Both dates are legitimate but describe different events. A metadata export must not overwrite one with the other or make the translation look contemporaneous. Check date fields and descriptive labels. Release check: Do not use translation creation date as source issue date.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 7: Visible signature

Illustrative source: “Handwritten signature visible; not independently authenticated.” Target-language meaning to retain: “The image contains a signature mark whose authenticity is not established by this translation.”

Reproducing a signature in an image is not certification. Translators may describe what they see while avoiding claims about the signer’s identity or intent. Inspect source capture and authentication records. Release check: Do not say ‘digitally verified’ without cryptographic evidence.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 8: Seal partly illegible

Illustrative source: “Official seal text partly unreadable.” Target-language meaning to retain: “Part of the seal’s inscription cannot be read confidently.”

A helpful translation can indicate which characters are unclear. Guessing a known institution’s name may obscure that the seal is damaged or different. Log the uncertain span and source crop. Release check: Do not complete the seal text from expectation.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 9: Transposed totals

Illustrative source: “Row total 1,042.50 in source.” Target-language meaning to retain: “The source total is 1,042.50 in the stated numeric convention.”

Thousands and decimal separators vary across locales. The numeric value must be preserved, and accounting interpretation checked with the source table. Reconcile digits, currency and calculation. Release check: Do not turn 1,042.50 into 1.04250 or 104250.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 10: Negative sign

Illustrative source: “Balance shown as −350.00.” Target-language meaning to retain: “The displayed balance is negative three hundred fifty in the source unit.”

A lost minus sign reverses an account’s apparent position. Pay close attention to dash glyphs and right-to-left text settings. Compare every signed amount and its table column. Release check: Do not replace a negative balance with a positive amount.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 11: Currency denomination

Illustrative source: “Amount stated as £12 in 1910 ledger.” Target-language meaning to retain: “The record states twelve pounds in its original historical context.”

An historical monetary denomination is not equivalent to a modern converted value. A conversion, if requested, is separate researched commentary. Verify symbol, date and ledger context. Release check: Do not silently modernise currency.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 12: Bilingual page order

Illustrative source: “Source French pages 1–4; English derivative pages 1–6.” Target-language meaning to retain: “The English translated rendition has six pages corresponding to four source pages.”

Different page counts are legitimate, but citations to the source must still identify the correct original folio and target location. Create a page alignment table. Release check: Do not reuse source page numbers as if they matched.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 13: Unembedded typeface

Illustrative source: “Font not embedded: missing glyphs on archival export.” Target-language meaning to retain: “The output lacks an embedded font needed for dependable text rendering.”

A file may display correctly on the creator’s computer using a locally installed typeface. That success does not establish an archival property. Check validator errors and font resources. Release check: Do not call the output fully self-contained.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 14: Glyph extraction error

Illustrative source: “Visible ‘μg’ copies as ‘mg’.” Target-language meaning to retain: “The rendered microgram symbol does not extract correctly as machine-readable text.”

This change multiplies a dose by a thousand if interpreted literally. A visible screenshot alone cannot validate character encoding. Test copy-and-paste, search and underlying Unicode mapping. Release check: Do not rely on visual inspection alone.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 15: Right-to-left segment

Illustrative source: “Arabic paragraph copied in wrong logical order.” Target-language meaning to retain: “The Arabic text renders but its extracted reading order is incorrect.”

A target PDF can be technically readable and still fail when searched or interpreted by software. Assess script shaping, punctuation and text extraction. Run native-language extraction and screen-reader tests. Release check: Do not claim full text integrity on appearance alone.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 16: Flattened note

Illustrative source: “Translator annotation flattened into image.” Target-language meaning to retain: “The explanatory note has become part of page artwork rather than selectable text.”

Flattening may protect a visual appearance while making commentary harder to search or access. Keep evidence text distinct from translated notes. Compare extraction and accessibility reports. Release check: Do not assume the flattened note meets the project’s reading requirements.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 17: Unverified source edition

Illustrative source: “Translated from scan; edition unconfirmed.” Target-language meaning to retain: “The translation used a scanned copy whose edition or revision is not verified.”

When an archive contains multiple revisions, translation may accidentally represent a superseded text. State the provenance limit until the curator resolves it. Check archival accession and edition metadata. Release check: Do not describe the source as definitive.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 18: Redaction failure

Illustrative source: “Confidential name blacked out visually only.” Target-language meaning to retain: “The name appears covered on screen, but secure removal of the underlying content has not been confirmed.”

A rectangle over text may leave the character data extractable or searchable. Redaction is an information-governance action, not a layout effect. Use approved redaction tooling and inspect extraction. Release check: Do not claim information was removed based on appearance.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 19: Digital signature after conversion

Illustrative source: “Signature invalid after PDF/A export.” Target-language meaning to retain: “The original signature verification no longer succeeds on the converted file.”

Changing signed bytes can invalidate a cryptographic digest. Maintain the signed original and document the derivative’s relationship instead of pretending the old verification applies. Verify source signature independently and preserve original. Release check: Do not duplicate the signature-valid claim onto the new file.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 20: XMP mismatch

Illustrative source: “XMP title differs from visible cover title.” Target-language meaning to retain: “The metadata title and page cover title are inconsistent.”

A user may locate the file through metadata and misunderstand which document was archived. The title can also relate to a different revision or language. Compare catalogue record, XMP and cover. Release check: Do not conceal discrepancy by changing one field without authority.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 21: Version zero ambiguity

Illustrative source: “Draft v0 marked ‘approved’ in filename.” Target-language meaning to retain: “The filename and in-document status conflict about approval.”

A title or filename is not decisive evidence of authorisation. In a translation, preserve the conflict until the document owner resolves source-control records. Check approval signature and release register. Release check: Do not choose the stronger status for convenience.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 22: Conflicting language tags

Illustrative source: “English text labelled French in PDF metadata.” Target-language meaning to retain: “The actual text language and metadata language label disagree.”

Language tags affect discovery and potentially assistive reading. A fluent translation can be hard to locate if the digital object advertises the wrong language. Inspect XMP and any structure language attributes. Release check: Do not declare language tagging correct solely because the text displays.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 23: Rotated table

Illustrative source: “Landscape annex loses rightmost column after export.” Target-language meaning to retain: “The archival conversion omitted a column from the wide table.”

Translation and repagination can affect bounding boxes. Missing columns may carry contractual conditions or balance totals. Compare source and target pages edge to edge. Release check: Do not approve a page based only on a text diff.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 24: Footnote marker lost

Illustrative source: “Clause 6 subject to note 3.” Target-language meaning to retain: “The qualification stated in footnote 3 applies to clause 6.”

A footnote can reverse the interpretation of an otherwise clear sentence. Keep its marker and relationship intact through reflow or repagination. Check cross-reference targets and note text. Release check: Do not publish a bare clause without its qualification.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 25: Obsolete hyperlink

Illustrative source: “External URL shown as supporting reference.” Target-language meaning to retain: “The document contains a reference URL, but future availability of that site is not guaranteed.”

PDF/A preservation does not ensure every external information source remains accessible. Reference capture is a separate institutional policy question. Archive the cited record if authorised and record access date. Release check: Do not say the PDF stores linked content when it does not.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 26: Embedded attachment claim

Illustrative source: “Appendix referenced; file attachment absent.” Target-language meaning to retain: “The text mentions an appendix, but no attachment is present in this PDF/A-1 derivative.”

PDF/A-1 has feature restrictions. A referenced appendix may need to be a separate managed archival object linked in repository metadata. Inspect original, records package and export report. Release check: Do not claim that a missing attachment is embedded.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 27: Colour legend altered

Illustrative source: “Red indicates rejection in source legend.” Target-language meaning to retain: “The source assigns the status ‘rejected’ to the red category.”

If colour changes in export, an apparently routine legend may imply the opposite category. Translate the written label and inspect actual rendered colour. Check legend, output intent and visual comparison. Release check: Do not infer decision from colour alone.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 28: Handwritten correction

Illustrative source: “Margin says ‘not approved’ in ink.” Target-language meaning to retain: “The handwritten marginal note states that approval was not given.”

Handwritten annotations may postdate the typed body and carry a different author’s decision. Preserve the medium and uncertainty when known. Review original handwriting and chronology. Release check: Do not combine handwriting into typed text without attribution.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 29: Correction record

Illustrative source: “2026 translation supersedes 2024 translation, not source.” Target-language meaning to retain: “A newer language rendition replaces an earlier translation while the original historical record remains the same.”

The word ‘supersedes’ requires an object: translation version or source document. Without it, the target can suggest the archive’s original changed. Check derivative relationship and correction reason. Release check: Do not suggest the historical document was revised.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 30: Preservation guarantee

Illustrative source: “PDF/A file format selected for long-term preservation.” Target-language meaning to retain: “PDF/A was chosen to support long-term preservation under repository practices.”

Format selection is not a guarantee of perpetual readability, retention, legal validity or institutional custody. Preservation includes storage, monitoring, integrity and governance. Check archive policy, fixity checks and migration plan. Release check: Do not say the file is ‘permanently preserved’.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 31: Accessible claim

Illustrative source: “PDF/A-1a validated; screen-reader review incomplete.” Target-language meaning to retain: “A file-format check for level A passed, but human accessibility testing is unfinished.”

Format conformance and usable semantic reading are different evidence streams. Translation can introduce new reading-order or language-tag problems. Review reading order and native-script output. Release check: Do not call the document fully accessible without testing.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Case 32: Public versus restricted

Illustrative source: “Archive master restricted; translated access copy public.” Target-language meaning to retain: “The protected master and the public translation have different access rules.”

PDF/A forbids encryption inside compliant files, but access can be managed at the repository layer. A public translation must still obey privacy rules. Confirm permissions and contents of the derivative. Release check: Do not assume PDF/A compliance grants publication rights.

Preserve the relationship between the source page and this statement in the derivative record. Record whether the statement is original content, a translator’s observation, a technical-format test result or an archival decision; these categories must never be merged in a formal summary.

Step-by-step translation and preservation QA

Create an archive/source inventory and a versioned extraction before writing target text. Document whether each page is born-digital, scanned or mixed, and where OCR is uncertain. Translate source meaning with terminology, page references and domain validation. Before PDF export, select a documented font strategy that covers required target glyphs and use a layout able to preserve tables and qualifiers. Export with the chosen PDF/A-1 conformance target and verify the resulting file in an appropriate validator. Inspect its warnings and errors rather than relying on a preset. Compare visible pages against the approved text and the original record, then test searching, copy-and-paste, text extraction and accessibility where relevant. Finally record metadata, a source relationship, reviewer approval, retention class and release date.

Keep security and evidentiary decisions separate from the format check. Do not remove historical signatures, access conditions or redaction requirements without authorisation. Protect the stored original and its checksum under repository controls. A human reviewer should test both the most sensitive character sequences and the most legally meaningful qualifiers. A cross-language reader should be able to answer: which record is original, what was actually translated, which parts were uncertain, who reviewed them, and what the validation result does and does not prove.

Frequently asked questions

Can I translate directly in the archived PDF/A-1?

Preserve the original under its records process. Produce a separately controlled derivative from approved source content, maintaining a link to the original and its integrity records.

Does PDF/A-1b make a file accessible?

No. Its main focus is dependable visual rendering. A requires additional structure-related features, but neither label guarantees meaningful accessible reading in practice.

Does PDF/A guarantee the translation is accurate?

No. The format standard and language quality are different. Conduct bilingual text comparison, domain review and provenance checks in addition to file validation.

Can I rely on OCR for names and dates?

Not without checking against source images, particularly for handwriting, diacritics, unusual typefaces and damaged archival material.

Do I need embedded fonts?

Font embedding is a core PDF/A preservation consideration, with limited technical exceptions. Use a validator and approved export process, and check all target-script glyphs.

Can PDF/A-1 contain an encrypted file?

Encryption is not permitted by the PDF/A-1 profile. Repositories may control access outside the file under their governance procedures.

Is an electronically signed original still signed after translation?

A derivative is not the original signed byte stream. Verify and retain the original; treat any signature or certification on the translation separately.

How should I write the archival metadata?

Record the derivative’s language, title, responsible creator, date and source relationship using the repository’s approved schema. Do not overwrite historical source facts.

What if a translated document has a different number of pages?

That is often acceptable if source-to-target page mapping and referenced clauses remain traceable. Revalidate layout, footnotes, annexes and citations.

Should I choose PDF/A-1a or PDF/A-1b?

Follow actual repository requirements and the capabilities of the approved workflow. Do not advertise one level based on a test for the other.

How do I prove PDF/A compliance?

Validate the final bytes with an appropriate tool and profile, retain its report and resolve failures. Supplement with visual, extraction and metadata checks.

What is the most important release record?

A document package linking original source ID, translator’s approved text, derivative version, file validation report, review approvals and distribution status.

A four-week implementation plan

Week 1 — source audit and extraction

Select two authorised archival records, preferably one born-digital and one scan. Inventory document identifiers, provenance, page order, attachments, signatures, language, revision and uncertain text. Keep original files untouched. Extract text and compare a sample against the image, marking ambiguous dates and names. Define what a translation derivative means in your repository. Build a small risk register rather than relying on memory or an untracked translator note.

Week 2 — faithful bilingual content

Translate the documents using a glossary and visible source references. Have another reader verify numbers, negation, historical labels, table headings, seals, units and legal status phrases. Keep annotations distinct from the source text. Create page alignment notes when line wraps or paragraph flow differ. Test at least ten potentially confusing characters in the actual target script, including hyphens, minus signs and accent marks. Finish only when each query has an owner and outcome.

Week 3 — PDF/A rendering and validation

Create a derivative PDF using the archive-approved export method and a clearly specified conformance level. Run a recognised PDF/A validator on the produced file. Check XMP fields, font embedding, glyph rendering, copy-and-paste, structural reading and page integrity. For scans, compare source image and extracted text separately. Review any flattening, colour or signature effects. The goal is two independent passes: technical format validation and human content verification.

Week 4 — release with durable provenance

Record the original archive ID and integrity information, approved source text, derivative title and language, translator or team, review decision, software validation report and intended access class. Release the versioned file through the proper repository process. Simulate correction of one mistranslated figure and document how a new derivative is created while preserving prior records. Check whether readers can locate both the authoritative original and the updated interpretation. A successful multilingual archive makes future reconstruction easier, not merely today’s visual output attractive.

Sources and ecosystem connections

Primary authority: ISO 19005-1:2005 official catalogue. Independent explanatory references: Library of Congress sustainability assessment of PDF/A-1 and PDF Association resources for ISO 19005. For binding clauses, corrigenda, legal retention and exact validation rules consult the applicable standard, archive policy and qualified specialists. The original examples above do not constitute legal or technical certification.

This specialist article refers readers to the protected eduKateSG Master Art of Translation (EDKSG-TRANS-MASTER-WORLD-000), the Vocabulary Learning Hub and How English Works: Translation and Meaning Across Languages. The architecture and canonical ownership of those larger guides remain unchanged.