How can a library, university or digital-humanities project translate a TEI P5 scholarly edition without changing what its manuscript witnesses actually say? Professional TEI XML localization is a problem of evidence, representation and interpretation. The reader may need a new-language introduction, contextual notes and interface labels, but the underlying transcription, variant readings and documentary identifiers must remain traceable to their sources. Translation is valuable precisely because it makes the evidence accessible without pretending that the translated explanation is itself the historical document.
People searching for TEI P5 translation, TEI XML localization, multilingual scholarly editions, critical apparatus translation or digital manuscript translation usually need more than an exported prose translation. They may require a source transcription, a parallel modern-language rendering, a machine-readable record of editorial responsibility, references back to page images and an apparatus of variants. If the translated result merges witnesses or silently changes the editorial reading, the scholarly record has been distorted.
This guide focuses on TEI P5’s documentary and editorial structures: the TEI header, text divisions, language tags, diplomatic transcription, normalization, critical apparatus, witness references, facsimiles, notes, entities, links, customization, validation and public presentation. It gives a full professional view of multilingual edition production rather than repeating a narrow quality-check rule about translating notes. The canonical reference is the Text Encoding Initiative’s continually maintained P5 Guidelines, including the critical apparatus and transcription modules.
Begin with the scholarly object, not with a text file
A TEI document can represent a printed book, manuscript, letter, inscription, oral history, dictionary, drama, correspondence archive or critical edition. Two TEI files that both contain XML and paragraphs may embody very different claims. A manuscript transcription records what a physical witness contains. An edited reading may resolve ambiguity. A translation offers a new linguistic rendering. A catalogue description explains the object. Good localization preserves these layers as separate scholarly propositions.
Before work begins, ask the edition owner what is being translated. Is it the author’s text, an editor’s introductory essay, metadata labels, a translation of the main text, witness descriptions, an exhibition interface or all of these? The distinction affects licensing, expertise and the permissible changes to the XML. A historical document may contain archaic spelling, uncertain readings and purposeful irregularity. An elegant target-language rendering is not a license to regularize the source witness in place.
Identify the authoritative source and edition policy
A scholarly edition needs an explicit source of truth. Record the TEI file version, responsible editors, repository location, associated page images, declared schema and any local encoding rules. Check whether the edition is diplomatic, normalized, eclectic, genetic or a parallel edition. These terms describe different relationships between source evidence and editorial intervention. A translator working without that policy can inadvertently make the most dangerous kind of mistake: a readable result that has lost its evidential distinction.
An edition may have competing witnesses, each with omissions, additions or variant spellings. The editor may choose a lemma for display while retaining alternative readings in an apparatus. Translation can explain those choices, but should not erase them. Establish a clear rule for which components are immutable evidence and which are target-language expressions. Save a source snapshot and a change ledger before modifying the target edition. Provenance is part of publication quality, not an administrative afterthought.
The TEI header is a record of responsibility
The TEI header provides documentation about the electronic text. It commonly includes file description, encoding description, profile description and revision history as appropriate to the document. A translator who treats the header as an unimportant preface misses information needed by future researchers. Titles, author statements, publication statements, sources, responsibilities, languages and edition histories determine how a digital text can be found, cited and interpreted.
Translate descriptive prose where a target-language audience benefits, but keep stable identifiers, bibliographic facts and institutional authority references intact. If a project introduces translated title variants, indicate how they relate to the original title rather than replacing the original record without explanation. Distinguish the historical author from the modern transcriber, editor, translator and technical encoder. Different people performed different intellectual acts, and a multilingual edition must leave those acts visible.
File descriptions and scholarly citations
Within the file description, a title statement can identify works and the people responsible for them. Publication information can identify publishers, dates and availability; source descriptions identify the manuscript or printed sources from which the electronic edition derives. When translating this prose, preserve the bibliographic identity of the source. A shelfmark is not an ordinary phrase. An archival reference or persistent identifier may be the only practical way for someone to verify the manuscript.
A target-language citation should remain unambiguous even if conventions for punctuation, ordering or abbreviated role labels differ. If a name has several established forms, retain the authoritative form and provide an alternate rendering with an explicit relationship. For example, translating a repository’s descriptive name may help a reader but should not silently replace its official institutional identity. Document which title is original, which is supplied by an editor and which is a modern translation.
Encoding description and local editorial conventions
A TEI encoding description can explain project practices: how abbreviations are expanded, how uncertain readings are marked, what normalization means, or which source features are retained. These statements are particularly important for translated editions because readers need to know whether a polished target text reflects the underlying source or an editorial decision. The localization of this explanatory prose should use precise terminology and avoid giving stronger assurance than the original policy provides.
Suppose the editorial policy marks illegible letters as uncertain but does not attempt reconstruction. A translation that says the text has been reconstructed would create a false methodological claim. Similarly, if punctuation in the edition is editorial, the target language should not present it as an original feature of the manuscript. Ask the responsible editor to review policy language, especially where translation has to distinguish transcription, transliteration, normalization, emendation, expansion and interpretation.
Language declarations are functional metadata
Language information in a TEI edition does more than announce that a text is in English, Arabic, Latin or another language. A document may contain several languages, scripts, quotations, glosses and translations. The language of a whole text is not necessarily the language of every embedded segment. TEI uses language tagging conventions such as xml:lang; appropriate language identifiers help processing, display, searching, text-to-speech and downstream analysis.
Use tags that reflect actual content, and remember that language, script and geographic variety are related but different attributes of a linguistic expression. A passage in Latin script may still belong to a particular non-English language; an Arabic-script quotation may occur in an English editorial note. Do not assign the target-language tag to unchanged source transcription merely because the surrounding edition is localized. Tagging decisions should be tested against how users search and read the new edition.
Source text and target rendering must remain distinguishable
A translation of a medieval passage may legitimately change syntax, idiom and word order, but it should not be stored as though it were a literal transcription of the witness. One safe publication design preserves the source transcription as an independent layer and adds the translation as a linked, separately identified layer. The reader can then move between what the witness attests and how the translator understands it. This is especially important where a single source expression permits multiple reasonable translations.
The edition should explain whether the target is literal, literary, pedagogical or analytic. A poetic translation may prioritize rhythm, while a diplomatic aid may preserve formal distinctions even at the cost of smoothness. Neither method automatically determines how the source XML should be edited. The encoded source remains evidence; the translation is a claim with an author, date and method. Treating those as different objects enables fair comparison, revision and proper citation.
Text divisions and reading order
TEI structures a text through elements such as front matter, body, back matter, divisions, paragraphs, heads and other specialized structures. The translation must preserve meaningful boundaries such as letters, chapters, speeches, stanzas, items and documentary sections. A passage translated into a longer target sentence may cross a punctuation boundary without needing to cross an encoded section boundary. Editorial decisions about segmentation affect navigation and citation.
Before translating, generate a map of stable IDs, division hierarchy and anchors used in citations. An online edition may link to a paragraph through an XML identifier, or allow a scholarly article to reference a specific line. If the translation rearranges content without a stable correspondence map, the original citation may point to the wrong passage. Decide whether the target edition shares source anchors, uses its own IDs linked to the source, or presents explicit segment alignment. Make the choice consistent.
Diplomatic transcription versus normalization
A diplomatic transcription attempts to represent selected features of a witness, including spelling, abbreviations, additions or deletions as defined by the edition policy. A normalized reading may modernize spelling or editorially regularize features to aid reading. The two are not interchangeable. When a manuscript uses an unusual spelling, the translator should not simply alter the transcription so that it resembles a modern target-language word. That confuses the source artifact with an interpretation of it.
TEI can encode alternatives such as original and regularized forms. These alternatives provide a way to display more than one reading without destroying the distinction. The editor may want visitors to switch between diplomatic and normalized views. A parallel translation should be designed to accompany those views, not overwrite them. Test a passage containing deliberate irregularity so that the localized interface explains the source’s status accurately and the viewer’s toggle remains honest.
Corrections, errors and editorial interventions
A source witness can contain mistakes, uncertain punctuation or apparent omissions. TEI can distinguish a source form considered erroneous from an editorial correction, and it can record who is responsible for the intervention. The purpose is not to conceal textual problems but to make them examinable. Translating a note that describes an editorial correction requires preserving the uncertainty and level of authority attached to that correction.
Consider a place name spelled inconsistently in two parts of a manuscript. An editor may propose a normalized place-name form and link it to a geographic authority. A translation may use an established modern exonym in explanatory prose, but it should not silently rewrite the actual manuscript spelling. When readers select a source view, they need to encounter the evidence. When they select an explanatory view, they need to know which spellings were supplied or regularized by the editor.
Abbreviations, expansions and uncertainty
Historical documents often use abbreviations that require specialized knowledge. A shortened form can have several possible expansions, especially when writing conventions have changed over centuries. TEI allows projects to distinguish an abbreviated source form from an expanded reading. A translator encountering such material should determine whether the translation is based on the expansion, the abbreviation, or an editor’s uncertain interpretation.
Uncertainty must travel with the interpretation. If the expansion is hypothetical, the translated presentation should not turn it into unqualified certainty. A footnote or tooltip may be the appropriate place to explain the uncertainty, rather than interrupting a readable translation every few words. The interface should also let qualified researchers inspect the source form. A reliable edition makes complexity accessible; it does not remove complexity by erasing the evidence that supports competing views.
Additions, deletions and revision history
Manuscripts may contain deleted words, inserted phrases, marginal additions, over-writing and corrections by more than one hand. TEI encoding can retain these layers when the project has the evidence and editorial policy to do so. Translation becomes more difficult because an addition is not just a phrase; it is an event in the document’s writing history. Flattening the latest reading into one paragraph may be useful for casual readers, but it cannot replace the edition’s representation of earlier states.
Decide whether the translation accompanies the earliest recoverable state, the final authorial state or a particular editorial reconstruction. Document that choice. If alternative states matter, the target display can provide a version switch and linked explanatory notes. A study of a novelist’s drafts may depend precisely on which sentence was crossed out and which was inserted. Accurate language is only one part of that story; the temporal order of revision is equally important.
Facsimiles, surfaces and zones
Digital manuscript editions often link transcription to photographs of pages or other documentary surfaces. TEI can describe facsimile resources and zones that correspond to writing areas, line regions or specific annotations. A translation that preserves the text but loses the image link deprives users of the means to inspect the source. If a paragraph was transcribed from a region near a torn edge, the visual location may explain a missing reading.
Test the relationship between a displayed target passage and the correct source image. A translated paragraph may contain more words than the line or zone it explains; that does not mean its geometry should be stretched to match the target prose. Keep source coordinates and image identifiers stable, while using an alignment layer for translation. A user should be able to click a translated sentence, see the relevant source region and understand any one-to-many correspondence.
Page breaks, line breaks and stable locations
A printed or manuscript source may have page and line breaks that support citation even when they are awkward for a flowing translation. TEI uses milestone-like structures and other markers to preserve such divisions. A reading interface may choose to hide visual page breaks while retaining navigable page anchors. Translators should not treat the absence of visible breaks in a target view as permission to discard the source location markers.
Check cases where a source sentence begins on one page and ends on the next. The target language may put the verb earlier, causing exact line-level alignment to become impossible. A sensible edition can preserve passage-level correspondence and mark the mapping as approximate when needed. Do not invent exact alignment merely to simplify interface development. The reader benefits more from honest correspondence than a falsely precise highlight that suggests a target word sits directly above an unrelated manuscript character.
The critical apparatus records alternatives, not noise
A critical apparatus can record variant readings across witnesses. In TEI, structures such as apparatus entries, lemma readings and alternate readings help editors describe the differences. Those terms are not decorative markup: they represent evidence of textual transmission. A translator who changes a lemma without examining how it relates to the witnesses risks falsifying the editor’s chosen base text or the alternatives available to readers.
Create a witness inventory with stable sigla and authoritative descriptions. For each apparatus entry, identify whether the target edition will translate the adopted reading only, summarize the variants, present full translated alternatives, or provide bilingual apparatus notes. The choice should follow an explicit editorial objective. Never merge two witnesses because their readings translate to the same word in the target language; distinct witnesses remain distinct sources even when their semantic outcomes seem similar.
Lemmas, readings and interpretation
A lemma is the reading selected or referenced in an apparatus according to the editorial method. A reading supplied by another witness may differ by spelling, word order, omission, morphology or meaning. Some variants are insignificant to an ordinary translation; others change an argument, a person’s name or a legal obligation. The edition should preserve the source-level distinctions even if the target prose can only represent them through notes.
For example, two manuscript witnesses might differ in a verb that could mean “commanded” versus “requested.” If the chosen lemma is translated as “commanded,” a target-language note can explain that another witness supports a softer reading. Such a note should not claim that the manuscript tradition is unanimous. Where translation neutralizes a grammatical distinction, the translator may need to retain a short philological explanation. The apparatus keeps the evidence, while the commentary helps readers understand why it matters.
Witness identifiers and documentary identity
Witness sigla identify particular manuscripts, editions or other documentary sources. They function much like stable database keys in the scholarly workflow. A localized interface may explain that a siglum refers to a manuscript in a named collection, but it should not casually rename the siglum throughout the encoded apparatus. Doing so risks breaking cross-references and making published scholarship difficult to follow.
Record manuscript shelfmarks, repository names, dates, material descriptions and version relationships with care. Distinguish the date of an original composition from the date of a surviving copy and from the publication date of a modern edition. A target-language sentence can easily collapse these into one “written in” claim if it is translated without context. Explicit source identity prevents a reader from confusing authorial text, scribal intervention and editorial reconstruction.
Editorial notes and translators’ notes are different acts
An editorial note may explain a paleographic reading, cross-reference a source or defend a textual decision. A translator’s note may explain why a target phrase was chosen, how an idiom was handled or why a distinction cannot be reproduced. Both are useful, but they have different authority. The language edition should preserve attribution and make it possible for the reader to tell which note came from the source editor and which was introduced by the translator.
An apparatus note that says “reading uncertain” is not the same as a translator’s preference between two synonyms. Keep the notes separately identified, even if they appear in one interface. If a target-language reader sees only one combined footnote list, use labels, authorship fields or accessible visual conventions to distinguish note categories. This safeguards scholarly accountability. Later editors can revise a translation without appearing to revise the original textual evidence.
Named people, places and events
Historical documents frequently mention people and places whose names changed over time or are known by several linguistic forms. The translator should distinguish the source spelling, established scholarly identification, modern conventional name and translated display label. Authority records or persistent identifiers can anchor these relationships. A word-for-word translation of a personal name is usually inappropriate, while an explanatory gloss may be essential for a target-language audience.
Where identification is disputed, preserve the uncertainty. If a source name might refer to two people, do not choose one merely because the target language has a familiar rendering. Likewise, a place whose borders or political designation changed should not be retroactively described with a modern label that alters historical meaning. A useful digital edition lets readers open a person or place record with evidence and variant names rather than hiding competing identities inside a single smooth sentence.
Dates, calendars and units of reference
Historical dates may use calendars, eras or local conventions unfamiliar to modern readers. Translation can explain the date, but conversion is a scholarly transformation that needs an identified calendar and method. A manuscript dated by regnal year, lunar month or local feast may not map to one uncontroversial modern calendar date. A target edition should retain the attested date expression and, if appropriate, provide a separately labelled normalized form.
The same care applies to measures, currencies, offices and geographic units. Translating a historical coin name as a modern currency name would imply equivalence that may not exist. Use glosses and contextual explanation instead of silently converting units. If an editorial project provides numerical conversions, it should cite its method and uncertainties. The translated language should not make the source appear more exact or more modern than the evidence permits.
Quotations and multiple embedded languages
A letter written in one language may quote another language, use set phrases in Latin or mix scripts within a sentence. A TEI edition can tag language changes at the passage or segment level. In a multilingual target edition, decide which quotations remain in the source language, which receive translations, and how each is marked. Preserve the distinction between a quotation authored in a different language and a translator-created rendering of that quotation.
Quotation marks, italic styling and script direction can be culturally variable, but those conventions should not obscure whether words are quoted, supplied or editorial. A mixed-direction passage may need careful Unicode bidirectional handling without inserting unrelated characters into source transcription. Test rendering and copying in realistic reading software. A historian searching for an exact source phrase should still find that phrase even after the interface language is changed.
Parallel text and alignment
Parallel editions allow readers to compare source and target passages. The underlying alignment may be by chapter, paragraph, sentence, verse, line or a custom unit. Do not assume one source sentence always corresponds to one target sentence: translations frequently combine, split or reorder units. Design a mapping that can represent one-to-many and many-to-one correspondences without implying literal word equivalence.
Choose stable anchors for both language layers. Link them through explicit relationships and ensure that new translations do not overwrite the anchors used by previous editions. Test navigation by selecting a source passage and viewing the correct target passage, then reverse the operation. A useful interface may highlight aligned paragraphs while explaining when the alignment is approximate. The educational value comes from making comparison possible, not from artificially forcing every language into identical segmentation.
Search and index behaviour across languages
Search in a scholarly edition may operate over transcribed source text, editorial normalization, translated prose, notes or all layers. Those are different search products. If a user searches for a target-language term, the system may find a translation that has no exact source-language cognate. The interface should show which layer matched. Otherwise readers may mistakenly cite a modern translated word as if it appeared in the historical manuscript.
Plan indexes for people, places, concepts and textual variants separately from raw full-text search. Transliteration and alternate spellings can improve discovery, but they should lead back to an authoritative record rather than create competing identities. Preserve source text for philological search and build target-language indexes as additional views. Test diacritics, Unicode normalization, script-specific tokenization and compounds with examples from the actual language pair.
Presentation is not the XML source of truth
A public edition may use TEI XML to generate HTML, PDF, EPUB or an interactive manuscript viewer. Each output has its own layout and accessibility issues. A polished website is not evidence that the underlying XML still parses or that its source references remain valid. Conversely, schema-valid XML does not guarantee that a target-language reader can understand the interface. Both representation and presentation require testing.
Keep transformations repeatable. Record stylesheet versions, language-specific display rules and output formats. When a translator changes a note, regenerate the target pages and inspect the affected links, tooltips and printed apparatus. A site can accidentally cache an old-language passage after the source XML has changed. Release checks should verify the actual public output as well as the stored source files, especially for bilingual toggles and deep links.
Custom schemas and ODD projects
TEI is designed for customization. A project may define an ODD specification and derive a schema to express its particular editorial practices. A translator cannot assume that every element used by one TEI project is allowed by another. Local conventions may specify how language tags, commentary or editorial responsibility are represented. Before editing a file, inspect the project’s schema, examples and documentation.
Do not add plausible-looking elements or attributes that the project has not defined merely because they would make translation easy. Instead, propose a documented encoding change and have the edition owner approve it. A change that appears harmless to one page may break indexing or publication transformations across the collection. Validation against the approved project schema is a necessary control, but reviewers must also check the editorial logic that no schema can fully express.
Validation: syntax and semantic integrity
The first technical test is that the XML parses correctly, uses expected namespaces and validates under the selected schema. Then test IDs, references, witness pointers, image links and language tags. Compare source and target inventories of protected elements. A successful validation report means the file satisfies formal constraints; it does not by itself prove that every translated claim is historically or linguistically sound.
Develop specific queries to detect broken references, missing language declarations, empty translator notes, orphaned images and accidental replacement of witness text. Keep a baseline for the counts and identities of apparatus entries, witnesses, divisions and facsimile regions. Some counts may legitimately change when a parallel translation layer is added, so establish permitted differences. Any change to a source witness should be treated as an edition revision requiring editorial authority, not a routine localization edit.
Worked case: a medieval charter edition
Imagine a charter preserved in two manuscript copies with different spellings of a place name and a disputed verb in a grant clause. The edition provides images, diplomatic transcriptions, a normalized reading and apparatus notes. A new-language edition needs an accessible introduction and a translation of the adopted reading. First establish the shelfmarks and witness sigla, then preserve the source transcriptions, image links and variant entries. Translate the introduction and explanatory notes under the editor’s approved terminology.
The crucial passage involves a verb that one witness renders as a command and another as a request. The translator must not erase that divergence by presenting one target-language phrase as the unanimous documentary reading. Use an explanatory note tied to the apparatus and retain the witness identifiers. Test that a reader can follow the translated grant clause to both page images and see where the competing readings occur. This workflow maintains accessibility without sacrificing source accountability.
Worked case: an author’s annotated draft
Consider a novelist’s draft with deletions, insertions, comments in the margin and a later fair copy. A translation commissioned for a museum exhibition may highlight how the author’s choices evolved. Here the revision sequence matters as much as the final words. The target-language display should explain each stage rather than simply present a smooth final translation. Keep the encoded additions and deletions linked to the correct surface and hand if those distinctions are supported by evidence.
Create translations of selected draft states with clear labels. Test a passage whose deletion changes a character’s motive, because that is where temporal flattening would alter interpretation. The accompanying note can explain why the earlier reading matters. Keep curatorial prose separate from authorial manuscript text and from the modern translator’s reconstruction. The final exhibition should make the creative process understandable while still allowing a specialist to inspect the documentary record.
Worked case: a multilingual oral-history archive
A collection of interviews may contain code-switching, uncertain audio segments, speakers using different dialects and interviewer interventions. TEI can help represent spoken text and editorial metadata, depending on the project’s encoding conventions. Translation should not erase speaker identities, timestamps, hesitations or the distinction between words that are clear and words that are uncertain. The target-language public transcript may be readable, but the source evidence must remain available according to consent and rights rules.
Start with permissions, anonymization decisions and the approved transcription. Then determine how the target language will represent names, interruptions and culturally specific expressions. If a statement was unintelligible in the recording, do not invent a fluent target sentence. Keep uncertainty visible and link the translated passage to the proper time span when the platform supports it. Validate rights as carefully as XML structure: a technically perfect multilingual edition is not permission to publish restricted testimony.
Human review needs several kinds of expertise
A literary translator may be skilled at prose but not at paleography. A historian may know the context but not the target-language style. An XML developer may ensure schema validity but not recognize a mistranslated technical office. Build a review plan with clear responsibilities. Subject experts inspect evidence-sensitive claims; language editors inspect clarity and register; technical editors test markup and rendering; rights specialists inspect publication constraints where relevant.
Reviewers should see the source and target in context, including image links and notes. A spreadsheet of extracted text often hides the difference between a witness reading and an editorial comment. Use issue records that include an element identifier, surrounding passage, proposed change, reason, reviewer and resolution. The goal is a documented decision process rather than an apparently unanimous final file produced by silently accepting one person’s choices.
Governance of corrections and new editions
When a factual transcription error is discovered during translation, do not quietly fix the protected source while producing a target-language version. Record it as a proposed correction to the source edition. The responsible editor can decide whether to issue an erratum, revise the transcription and update any derived languages. Separating correction from translation protects the edition’s scholarly chain of custody. It also avoids creating language versions with unexplained differences.
Versioning should distinguish a source transcription release, an editorial correction release and a translation revision. The translation may be updated for style without changing source evidence; conversely, a corrected source reading may require retranslation of an affected paragraph and apparatus note. Preserve earlier published versions if cited and provide clear edition dates. Researchers need to know which textual state supported an argument, even if the website always displays the most recent version by default.
Frequently asked questions: should manuscript text be translated in place?
Usually not when the encoded content represents the source witness. A translated edition should retain the witness’s transcription and add a clearly identified language rendering or a parallel text layer. Translating in place would make it difficult to verify what the original source said and could break apparatus references. A standalone translation without the full TEI source can still be useful, but it should carry citations back to the source edition and name its own editorial method.
The right architecture depends on use. A student-facing excerpt needs less elaborate machinery than a critical edition used for textual scholarship. Both should tell readers where the source ends and modern interpretation begins. That distinction is the central principle of ethical, intelligible TEI localization.
Frequently asked questions: can automatic translation handle TEI?
Automatic tools may help with selected editorial prose after the system has separated eligible language from identifiers, source transcription and markup. They do not independently establish the historical meaning of an uncertain reading, which witness an editor should prefer, or whether a person is correctly identified. A safe workflow protects evidence fields, extracts approved translation units, retains context and requires qualified review.
Any automated round trip should validate syntax, references and protected content. Linguistic review should inspect the source image or reliable transcript where interpretation matters. Speed is not a substitute for provenance, and an XML file that still opens may nonetheless contain incorrect historical assertions. The most responsible use of automation is to reduce routine handling while keeping scholarly decisions visible and reviewable.
Frequently asked questions: what is the difference between TEI and a word-processing document?
A word-processing document can show styled paragraphs and footnotes. TEI encodes the nature and relationship of content: a witness reading, an editorial correction, a manuscript zone, a named person or a revision state can be represented as structured information. That structure enables different displays, searches and scholarly analysis. Translating a TEI edition must therefore preserve conceptual relationships that would be invisible in a flat text export.
A simple DOCX version may still be useful as an editorial review artifact. It should not become the only master if the project depends on machine-readable witness links and textual variation. Check the reviewed prose back into the appropriate TEI layer, validate it, and regenerate the public presentation rather than treating the temporary review file as the edition’s authoritative source.
Final release checklist for a multilingual edition
Confirm that the authoritative source version is recorded; witness text and sigla remain intact; protected identifiers are unchanged; translated prose has correct language declarations and attribution; apparatus entries still point to their source readings; images and zones open correctly; divisions and citation anchors work; source-to-target alignment is traceable; public display and search distinguish language layers; project schema validation passes; and editorial, linguistic and technical reviewers have approved their respective evidence.
Also confirm accessibility, rights and preservation planning. A reader should be able to tell what was attested, what was edited, what was translated and what remains uncertain. A future scholar should be able to reconstruct the relation between the target edition and its source. Those are not optional extras: they are the conditions under which translation enlarges access without replacing evidence with an untraceable paraphrase.
References and related reading
Start with the TEI P5 Guidelines, including their guidance on critical apparatus and element descriptions for readings. For a narrower operational check of editorial note localization, see TEI editorial-note translation QA. For adjacent but different structures, JATS journal XML and BITS book XML address publishing interchange rather than manuscript witnesses and textual criticism.