How do people translate TEI P5 editorial notes quickly without altering the historical text scholars are studying? Separate the source witness from modern editorial explanation, localize the commentary in its proper language layer, and verify every identifier, reference and transcription boundary. The quickest dependable workflow works on meaningful scholarly units, not a blind list of every text node inside an XML file. It protects the evidence and improves access without implying that the source artifact itself has changed.
The Text Encoding Initiative’s P5 Guidelines support structured scholarly texts, annotations, variants, pointers and descriptive metadata. In one edition, a <note> might explain an obsolete expression; a <seg> might mark a phrase; a page-break element might align transcription with an image; and a link might identify a passage discussed in another document. Those structures carry a history of editorial decisions. The translator must understand which sentences are contemporary explanation and which are readings, quotations or evidence.
Consider a digitized nineteenth-century school register used in a museum exhibition. The edition preserves spelling from the manuscript but contains English notes explaining administrative terms, uncertain names and the origin of the collection. A French edition should make those notes accessible while preserving the register’s recorded entries, uncertainty and page references. This manual shows how to manage that distinction, how to diagnose failures and how to transfer the method to a very different kind of digital edition.
1. Define the edition’s scholarly promise
Before extracting a single sentence, establish what the edition claims to represent. A diplomatic transcription seeks to preserve features of a particular witness; a reading edition may normalize some spellings; a scholarly apparatus compares evidence or records editorial interventions. Translation must not cause the project to switch quietly between these models. Readers deserve to know whether they are looking at historical wording, a modern interpretation or a translated explanation.
Write an editorial brief that names the source witness, the intended language audience, the permitted changes and the forbidden changes. For the school register, an English note describing an obsolete classroom rank may be translated, but an original name or date recorded on a manuscript page normally is not. A separate modern-language paraphrase can be valuable if visibly identified as interpretation rather than substituted for the diplomatic reading.
The brief should also assign responsibility for difficult terms, uncertain readings and cultural explanations. A literary translator may be skilled at elegant target prose yet lack access to the paleographic decision that produced a particular reading. A subject editor must resolve that ambiguity. Translation becomes faster when those ownership boundaries are explicit at the beginning rather than argued over during final proofreading.
2. Make a map of the TEI document before touching content
TEI is XML, but not every text node is a sentence for translation. The header records metadata and editorial decisions, the text body contains source material and structural markup, and additional mechanisms may link passages or hold annotations. A translator-facing inventory should identify the element, its context, language, relevant xml:id and link targets, and whether the node is part of the protected witness.
A raw XML export full of element names loses this context. A phrase inside <note> may be a modern explanation, a quotation from a source, or a citation. A phrase inside <seg> may be an actual segment of manuscript text. Treat the element name as a clue, not a final verdict. Editorial policy and local encoding conventions determine the authorized changes for that particular edition.
Create a selection rule that errs on the side of restraint. Extract approved editorial prose and descriptive fields, preserve protected source transcriptions and pass ambiguous nodes to an editor. Include enough surrounding context for translators to identify what a note explains. Without this context, a short translation of “the second hand” could refer either to handwriting by a later scribe or to an anatomical description.
3. Distinguish the historical witness from modern editorial commentary
The source witness is the book, manuscript, inscription or document being represented. Its wording is evidence even when a modern reader would prefer clearer grammar. Editorial commentary is a later explanation about that evidence. A translation workflow must keep these categories visibly separate, particularly when a note contains a quotation followed by interpretive discussion in the same paragraph.
For example, a register entry might spell a pupil’s surname inconsistently across pages. A translator should not silently harmonize the surname because the result looks more professional. The note can explain that the two forms probably refer to the same person, with uncertainty stated explicitly. Preserve the forms the edition transcribes and the evidence the editor used to link them.
If the project includes an actual translation of the historical source, that should be represented as a clearly identified separate translation layer or edition with its own provenance. Do not replace an original witness segment with target prose and leave the metadata pretending it is still a transcription. Scholars depend on knowing where evidence ends and interpretation begins.
4. Use xml:lang as a statement about content, not appearance
The TEI global attribute xml:lang can describe the language of encoded content. It is not a machine-translation command and it does not guarantee that readers will see a language menu. Mark language where the edition’s schema and policy permit it, and ensure the annotation actually contains the language indicated. A stale English language tag on a translated French note misleads tools that index or select text by language.
Some passages contain quoted Latin phrases, names in another script or bilingual remarks. A single language label for an entire document may not describe every embedded span. Determine the right granularity with the edition’s language policy, and keep scholarly quotations marked as quotations rather than normalizing them into target-language prose. Language metadata should illuminate the intellectual structure instead of flattening it.
When updating language declarations or multilingual descriptors, inspect the TEI header and any project-specific indexing configuration. Some systems derive language choices from the header, while others inspect individual elements. A valid language tag with a missing or incompatible presentation layer may still make the translation undiscoverable. Language tagging and interface behavior require separate tests.
5. Preserve xml:id and the network of pointers
An xml:id identifies an element so that another element or document can refer to it. Its value may contain recognizable words, but it is not a label to translate. A note referring to #folio-12 expects that target to remain resolvable. Replacing the identifier with a translated phrase while leaving inbound links untouched breaks scholarly navigation even when the XML remains well formed.
Treat pointers as a graph of dependencies. Before translation, collect target IDs and the elements that refer to them. After translation, confirm that every local fragment reference still resolves and that no unintended duplicate IDs appeared. When editors genuinely restructure the edition, they need a migration plan for links, citations and external references. Routine translation should not perform that restructuring implicitly.
Check both directions: does every note point to the intended passage, and does a reader examining a passage see the correct note in the published interface? A link can be syntactically valid yet lead to the wrong nearby segment after a mistaken copy-and-paste. Automated resolution checks and visual scholarly review solve different parts of this problem.
6. Translate notes according to their function
TEI notes may be explanatory, textual, bibliographic, linguistic or editorial. A textual note recording an alternative manuscript reading is not interchangeable with an exhibition note explaining why the artifact matters to visitors. Inventory note types where the edition provides them, and review the language requirements for each. Some notes should retain a quoted historical expression while translating only the words that explain its meaning.
A note about an obsolete school rank may need one familiar target-language description followed by a precise historical qualification. Replacing the rank with the closest modern job title without explanation risks anachronism. A note documenting a scribal erasure requires different treatment: the translator should describe the evidence and level of certainty, not embellish it into a confident story.
Establish a review protocol that asks what the note claims, what evidence supports it and whether the target text preserves uncertainty. A strong translation makes a scholarly claim intelligible while avoiding a claim stronger than the editor’s source. For a complex note, a bilingual reviewer should inspect the source document or a facsimile as well as the extracted sentence.
An illustrative witness-and-note fragment
This self-contained example is well-formed XML and shows the conceptual boundary between preserved witness text and a French editorial note. It has not been validated against a specific TEI ODD customization. A real edition must also obey the project’s schema and publication conventions.
<TEI xmlns="http://www.tei-c.org/ns/1.0">
<teiHeader>
<fileDesc>
<titleStmt><title>A School Register: Editorial Example</title></titleStmt>
<publicationStmt><p>Illustrative training fragment.</p></publicationStmt>
<sourceDesc><p>Fictitious classroom register.</p></sourceDesc>
</fileDesc>
</teiHeader>
<text><body><p>
<seg xml:id="entry-1" xml:lang="en">Class the second.</seg>
<note target="#entry-1" xml:lang="fr">Le classement historique demande une explication contextuelle.</note>
</p></body></text>
</TEI>
7. Guard documentary quotations and diplomatic spellings
Editorial notes sometimes quote the witness directly. A translation engine may regard the quotation as ordinary text and modernize spelling or punctuation, thereby destroying evidence. Mark quoted source spans as protected or supply explicit instructions that preserve them exactly. The surrounding interpretation may change language, but the quotation remains a record of what the original artifact appears to say.
If the edition’s policy calls for a separate translation of the quotation, present it as a distinct translated rendering and maintain access to the original reading. A bilingual audience can benefit from both, particularly where an obsolete word, dialect form or scribal abbreviation does not have a neat counterpart. The mistake to avoid is making a modern gloss look as though it was the original manuscript phrase.
Run a protected-token comparison on all diplomatic quotations, dates and identifiers. Then have a scholar review a selection of ambiguous passages. A character-level exact comparison catches unauthorized changes, but only expert reading can catch a gloss that is accurate in isolation yet distorts the historical implications of the quotation.
8. Treat facsimile links and page breaks as evidence anchors
Digital editions often connect transcribed passages to images of pages or folios. That alignment allows readers to check a reading against the physical witness. A translator might be tempted to rearrange paragraphs to improve narrative flow, but the resulting page correspondence could become inaccurate. The translation should preserve the ability to navigate from commentary to the particular area of evidence being discussed.
A page-break marker is not a decorative separator. If a note says an ink change appears after a folio boundary, moving that boundary changes the claim’s location. Keep facsimile identifiers, image URLs and page/folio references under controlled editing. Translate an explanatory label such as “rear side of the leaf” only after confirming that the target term refers to the correct physical side.
Test a sample of links in a real edition viewer. Confirm the correct image loads and that the relevant manuscript detail is visible, not merely that the URL returns a successful response. Where image rights restrict access, record the limitation without inventing a complete visual QA pass. The integrity of the scholarly claim depends on access to the correct evidence.
9. Keep textual variants distinct from explanatory translations
An apparatus may document several readings of the same location across different witnesses. Those variants are evidence, not competing translations proposed by a language team. A reader needs to know which source supports each form and what editorial selection has been made. Translating the prose around a variant must not collapse multiple readings into a single smooth sentence.
The TEI apparatus vocabulary, including elements such as <app>, <lem> and <rdg>, requires particular care when used in an edition. A lemma can identify the reading represented as primary under the project’s editorial method, while readings can identify alternatives and their witnesses. The precise encoding depends on project practice and customization, so inspect the actual schema before making assumptions about permissible nesting.
A bilingual explanatory note can help readers understand why two witnesses differ, but it must not change which witness contributed which form. Compare variant identifiers and witness pointers before and after translation. Then check that the displayed apparatus still attributes every reading correctly. A polished target-language paragraph is not an adequate substitute for a faithful chain of evidence.
10. Preserve uncertainty, editorial responsibility and provenance
Scholars frequently express qualified judgments: an attribution may be probable, a damaged letter may be uncertain, or a date may be inferred from context. Translation can accidentally transform this caution into certainty. A target sentence that says “the teacher wrote” when the source says “possibly written by a teacher” is a factual alteration, not a stylistic improvement.
Review words that convey degree of confidence alongside any structured certainty or responsibility attributes used by the edition. They are complementary. A prose qualification and a machine-readable confidence indicator should not disagree. The editor who made the original reading may not be the translator who explains it, and the publication should retain the distinction where provenance matters.
Write a decision note for every significant uncertainty that cannot be carried directly into the target language. Describe the evidence, alternative readings and rationale for the chosen wording. The purpose is not to make readers read a pile of production notes, but to ensure that future editors can reconstruct why the localized edition uses a particular qualification.
11. Treat names, dates and titles as scholarly identifiers
Personal names and place names are often more complex than they appear. The source may contain several spelling variants, while a modern authority file may assign a standardized form. Translating a surrounding note does not authorize replacing historical spellings with an authority label. Keep source transcription, normalized identification and explanatory prose in their appropriate roles.
Dates demand equal care. A source may use an older calendar convention, uncertain year or academic term. Translate the explanation of how the date is understood without silently converting it into a modern value unless that editorial conversion is explicitly intended and documented. A school register’s term name may not map cleanly onto a current school calendar or educational level.
For titles, distinguish the title supplied by the historical document from the modern title assigned by a cataloger. A translated display title can be useful for discovery, but the original title should remain available where relevant. Verify that citations and bibliographic identifiers still point to the same work rather than accidentally creating the appearance of a different source.
12. Record terminology decisions about historical concepts
Historical terminology resists simple substitution because institutions, ranks, materials and legal categories change over time. A school official’s position in one century may not match any modern equivalent in the target culture. Build a concept entry with the original form, its period-specific meaning, competing interpretations and the explanation chosen for the reader. That record is more useful than a one-word bilingual list.
Ask whether the target term imports a misleading present-day association. If it does, consider a descriptive translation or a retained original term with a compact explanation. This is particularly valuable for education, where seemingly familiar words such as class, form, college and master have carried different institutional meanings. The goal is historical clarity, not immediate familiarity at any cost.
Apply the decision consistently to notes, introductions, catalogs and captioning, but allow differences of function. A concise figure label may use a shortened phrase that is expanded in the main note. Record the relationship so editors know that the two forms refer to one concept rather than assuming they are unrelated terms.
13. Handle cross-language alignment as a separate editorial layer
A project may want the reader to compare an original segment with an authorized translation or with parallel commentary. Alignment links must communicate those correspondences explicitly. Do not fake alignment by making the target edition’s paragraph order look approximately similar. Historical language, sentence length and line breaks seldom map one-to-one in the way a naive spreadsheet expects.
TEI includes mechanisms for segmentation, alignment and stand-off linking. Choose a model that fits the edition’s scholarly promise and local customization. A short phrase in the original may require several clauses in the translation, and two source segments may reasonably correspond to one target segment. Represent that relationship deliberately, with clear boundaries and provenance, instead of forcing a false symmetrical segmentation.
Test alignment in a reader-facing interface as well as in XML. Can a reader click the target-language explanation and reach the intended passage? Can an editor trace it back to its source? A link graph that validates but highlights the wrong sentence is a scholarly defect, and the repair belongs to the alignment mapping rather than to a wholesale rewrite of the historical transcription.
14. Avoid inserting translations into protected transcription nodes
An editorial system may store transcription in <p>, <l>, <seg> or more specialized structures. The element alone does not tell you whether it is safe to translate. A language service that changes every text node in a <p> can erase original spelling, conflate witness evidence and invalidate established citations. Define localization boundaries from edition policy and structural context rather than generic XML element names.
For a register, leave the source entries in their designated transcription layer and add target-language explanations using the project’s supported editorial mechanisms. If translated source text is part of the deliverable, identify it clearly as a parallel rendering and provide a means to compare it with the witness. Do not substitute the rendering and keep the label “transcription” merely because the output looks clean.
A protected-node report is a practical safeguard. It lists the source witness fragments that were excluded from translation and compares their hashes or normalized text across editions. Hashes alone cannot test semantic fidelity, but they are excellent at detecting accidental edits to text that was not authorized to change.
15. Make TEI header changes explicit and minimal
The TEI header can record the title, publication information, source description, language usage, editorial practice and revision history. A localized edition may need a translated display title or additional language information, but an automated pass over all header text would be reckless. Bibliographic data often contains original titles, publication statements and cataloging references whose identity must remain stable.
Decide whether the translation represents an additional language view of the same edition or a separately issued scholarly publication. That distinction affects citations, identifiers, revision records and edition statements. A new language edition should not silently inherit claims of review or publication that applied only to the source edition. Likewise, an existing witness identifier must not be treated as an optional translated title.
Before release, compare the source and target headers field by field. Document authorized localized fields and any deliberate changes in edition metadata. A librarian or editor should check that the target publication points back to the correct source and that future researchers can distinguish a translation of commentary from a newly transcribed historical document.
16. Validate XML structure and TEI customization separately
A file can be well-formed XML while violating its TEI schema or the edition’s local customization. Well-formedness means markup is syntactically organized, but it does not prove that an element belongs in a particular place or that a pointer is meaningful. Use the schema or ODD customization for the actual edition, not an arbitrary generic validator, and preserve namespace declarations during transformations.
A translation process should leave element order and nesting intact unless an editorial change has been approved. A long translated note may tempt someone to insert a new paragraph at a location the customization does not allow. Fix the model appropriately through editorial approval or restructure the note within allowed content, rather than hiding validation errors under an apparently valid XML wrapper.
Report validation layers honestly: XML parsing, local schema conformance, ID and pointer resolution, viewer rendering and scholarly meaning review. Each answers a different question. A tool passing the first two cannot confirm that an explanation accurately represents an inscription or that a facsimile pointer highlights the right line on a scanned page.
17. Review the rendered apparatus and its reading order
A digital edition is experienced through its interface: notes may appear in a side panel, footnote list, popup or aligned bilingual columns. Longer translated text can overflow a panel or obscure the witness line that it explains. If numbering changes or footnotes are reordered, readers may attribute commentary to the wrong passage even though the underlying identifiers survive intact.
Inspect representative pages with dense notes, variant readings, embedded quotations, images and mixed writing directions. Compare the source and target editions’ navigational affordances, not just their typography. A bilingual reviewer should follow a citation from a paragraph to its note and then to the facsimile. That small realistic journey reveals mistakes that a static screenshot cannot show.
Where an interface cannot display all language variants simultaneously, document what is available and offer a clear navigation route. Accessibility matters here too: the relationship between a note and its anchor should remain understandable by keyboard and assistive technology users. A visually elegant alignment is not sufficient if the scholarly structure becomes inaccessible.
18. Diagnose errors by their broken scholarly relationship
When a translated note points at the wrong passage, test the pointer before rewriting the note. When a quoted reading differs, compare protected transcription nodes. When a language selector misses a note, inspect language metadata and the publication pipeline. When a historic term sounds modern and misleading, return to the glossary and the editorial evidence. These symptoms arise at different layers, and mixing repairs can introduce new faults.
Keep a defect record that identifies the source witness, affected element ID, link, language, scholarly consequence, reproducing action and approved correction. This allows translators and editors to distinguish a linguistic defect from a structural one. A valid cross-reference leading to the wrong witness is especially serious because readers may trust the neat interface without discovering that the supporting evidence has shifted.
After every repair, rerun the relevant structural test and the reader journey that revealed the problem. If the changed note is linked to several places, review all of them. Aim for small, traceable corrections. A rushed global rename can turn one broken pointer into dozens and make later scholarly auditing far more difficult.
19. Transfer exercise: translate notes in a poetry manuscript edition
Now imagine the next project is not a school register but a manuscript poetry edition with three witnesses, line breaks, uncertain spellings and an apparatus documenting alternative readings. The target readers need clear modern-language commentary, but the poetic source text and all witness-specific variants must remain evidence. One note includes a quotation from the poem, followed by a hypothesis about its meter and a pointer to a folio image.
A defensible translation preserves the poetic quotation and witness readings, translates the explanatory hypothesis with its original uncertainty, keeps the note’s target and image reference, and verifies alignment with the correct line. It records a terminology choice for meter if the target culture uses a different conventional label. If the edition publishes a translation of the poem itself, that becomes an explicitly identified separate rendering, not a replacement transcription.
The answer illustrates why a general instruction to translate all notes is insufficient. One note can contain protected evidence, localizable interpretation and machine-readable relationships. Build a three-part extraction record for this note showing each category and its reviewer, then test the rendered apparatus. If a reader cannot distinguish the original verse from the translator’s commentary, the edition fails its scholarly responsibility.
20. Create a release trail that future editors can follow
A completed translation package should record the source edition revision, local TEI schema or ODD, target-language policy, translation decisions and which nodes were protected. Provide a machine-readable diff of changed elements and an editorial summary explaining the few meaningful changes. A person reviewing the edition years later should be able to discover whether a phrase reflects the witness, the original editor or the translator.
Before distribution, validate XML, check project schema constraints, resolve local pointers and test representative facsimile links in the publication interface. Ask a scholar to inspect significant claims and uncertainty statements, and a native-language reviewer to assess readability. Confirm that citations, header metadata, original titles and source descriptions still identify the same material. Document anything that could not be tested.
Release the translated commentary with clear provenance, not merely with a new language label. Good scholarly translation invites broader readership while preserving the evidential chain. It should make the artifact easier to understand and just as easy to check. That is the practical meaning of translating quickly without rewriting history.
21. Design a translation memory that respects editorial boundaries
Translation memory can accelerate repetitive explanatory material, but it becomes dangerous when it mixes witness transcriptions with editorial notes. A recurring historical phrase may appear in a citation, a diplomatic transcript and a modern description; reusing one target translation across all three destroys distinctions that the edition was built to preserve. Segment the memory by editorial role, not merely by XML element name, and retain the source location and edition revision for each approved reusable segment.
Include enough commentary context to disambiguate short sentences. The phrase “the hand changes here” may refer to a different scribe, not to someone’s physical hand. A note explaining a page tear may use “loss” in a material rather than legal sense. A low-context translation memory will suggest fluent but incorrect phrases. Flag uncertain matches for a subject editor, and make the accepted wording traceable to its supporting passage.
Do not automatically propagate an old translation after an editorial note has changed its scholarly claim. Compare the source revision, confidence language and cited evidence before reusing a segment. Translation memory is a retrieval aid, not the owner of historical interpretation. Its value comes from reducing needless retyping while still allowing each note to carry exactly the meaning warranted by its evidence.
22. Check discovery, indexing and multilingual citations
A scholarly edition can be structurally correct yet practically invisible to its intended target-language readers. Search systems may index only one language, omit notes stored separately, or construct snippets without showing whether a sentence is transcription or editorial commentary. Verify how the publication indexes translated descriptions, titles and notes, and whether results clearly distinguish those categories. A translated catalog summary should guide readers to the correct witness rather than imply that the witness was composed in the target language.
Test likely reader searches using both an original proper name and a target-language descriptive phrase. Confirm the result leads to the right document and, when possible, the right note or passage. Language-specific search stemming may improve retrieval, but should not silently normalize exact archival identifiers. A source ID is a stable citation target even when its visible explanatory title differs across languages.
For citation export, check what the system supplies as the work title, editor, translator, language, source witness and publication edition. A translated webpage title is not automatically a different historical work. Where the editorial project provides a recommended citation, localize explanatory words carefully while preserving identifiers and bibliographic responsibility. This helps prevent a widely shared translated note from losing its connection to the evidence it explains.
23. Preserve the chronology of editorial interventions
A digital edition may accumulate changes by several hands across years: initial transcription, corrected readings, newly digitized pages, translated notes and later scholarly revisions. Translation is one part of that chronology. Avoid editing revision records as though they were plain descriptive paragraphs. Each record may explain when the evidence was re-examined or who made a correction, and later readers need that sequence to judge reliability.
When an old note is corrected, distinguish a correction to the scholarly claim from an update to the translated wording. The first may require review of every language edition; the second may concern only target-language clarity. Record the revision relationship so a future editor can tell whether parallel notes have been checked against the same source evidence. A clean-looking publication with undocumented divergence between languages creates a difficult research problem.
Before handover, compare the revision history, current editorial decisions and translation change log. Confirm that all target-language qualifications reflect the latest authorized source text and that no unreviewed historical claim has entered through an old memory suggestion. Preserve the chronology as provenance, not as a decorative footnote. That discipline allows the edition to develop without obscuring who changed what, why, and on what evidence.
Practical review checklist
Confirm the edition model, protected witness boundaries, permitted commentary fields and intended audience. Inventory xml:id, pointers, facsimile references, quotations, variants, uncertainty markers and language declarations. Check names, dates, historical concepts, diplomatic spellings and authorized notes against editorial policy. Require a subject expert for ambiguity and a target-language reviewer for clarity.
Then parse XML, validate against the actual TEI customization, check IDs and links, render the edition, and test navigation from notes to evidence. Record each changed editorial element, the source revision and any remaining limitations. Avoid changing manuscript transcriptions, identities, date claims or existing content outside the approved localization scope. A checked file is not a published scholarly edition until the publication process has independently been authorized and completed.
Limits and summary
TEI is expressive enough to support different editorial architectures, so no generic note translation rule can certify every digital edition. Local schemas, audience conventions, provenance practice and rendering platforms must be considered. Some historical ambiguities have no perfect target-language equivalent, and a reviewer should preserve that uncertainty rather than pretending it disappears through a fluent phrase.
The dependable method protects the source witness, translates editorial explanation in its declared role, and validates the link between the two. Readers receive accessible commentary without losing the means to verify it. Precision is not an obstacle to speed; it is what prevents a small translation task from becoming an untraceable rewriting of historical evidence.
Frequently asked questions
1. Is every TEI note safe to translate?
No. A note may contain modern commentary, source quotations, textual variants or bibliographic data. Classify the material within the note before editing it.
2. Should the historical source transcription be translated in place?
Generally not. If a translated rendering of the source is commissioned, publish it as a clearly identified translation layer or edition with preserved access to the witness.
3. Why keep xml:id unchanged?
IDs are reference handles used by notes, links and other tools. Translating them can break citations and navigation even when the file remains syntactically valid.
4. Does xml:lang automatically create a language selector?
No. It identifies language for encoded content; the publishing platform needs its own supported mechanism to present or select languages.
5. Can a translated note change the wording of a quoted manuscript phrase?
Only if an explicitly separated translation is the deliverable. A diplomatic quotation should remain the quoted evidence rather than being silently modernized.
6. What does TEI schema validation prove?
It tests conformance to structural constraints. It does not establish historical truth, citation accuracy, correct facsimile alignment or acceptable target-language interpretation.
7. How do I handle an uncertain historical title?
Preserve the evidence and its uncertainty; offer a descriptive target-language explanation where helpful. Avoid choosing a falsely precise modern equivalent.
8. What if a note links to the wrong scanned folio?
Treat it as an evidence-link defect. Check the element ID, pointer target and viewer mapping before changing note prose or manuscript transcription.
9. Can separate witnesses be merged to simplify translation?
Not without a scholarly editorial decision. Witness distinctions and variant attribution are core evidence in a critical edition and should remain traceable.
10. What should a translation handover contain?
The source revision, TEI customization, authorized changed elements, protected-node checks, term decisions, ID/link validation, rendered QA and unresolved questions.
Authoritative sources
- TEI P5 Guidelines — Linking, Segmentation, and Alignment — official TEI P5, version 4.12.0 dated 28 July 2026.
- TEI P5 reference — note — attributes and semantics for note structures.
- TEI P5 reference — seg — segmentation and global attributes.
- TEI Consortium — organization maintaining the Guidelines.