VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

Top Ways to Translate Correctly | TEI P5 Parallel Text Alignment and Manuscript Notes

Translating TEI P5 parallel texts correctly means rendering a source passage into a new language while retaining its historical witness, manuscript uncertainty, encoding structure and explicit relationship to the original text. In TEI XML translation, the crucial questions concern xml:lang language identification, xml:id references, linkGrp alignment, editorial notes, facsimile pointers and the difference between transcription and interpretation. This guide explains those decisions as an applied method for digital humanities, scholarly editions, libraries and historical archives.

The visible problem is a translation that sounds fluent while silently erasing evidence: illegible letters become confident facts, a later hand becomes the original author, two aligned passages lose their reference links, or a folio page no longer connects to its image. Such errors cannot be repaired merely by choosing a better synonym. A responsible TEI translation checks what the witness says, what the editor infers and how the machine-readable file keeps those layers distinct.

An immediate checklist for TEI P5 parallel-text alignment, manuscript note translation and scholarly XML review:

  • Identify the original witness and sourceDesc.
  • Protect xml:id and URI pointer targets.
  • Mark real language changes with xml:lang.
  • Represent one-to-many correspondence honestly.
  • Retain uncertainty, additions, deletions and corrections.
  • Check facsimile links and editorial responsibility.

The canonical job and boundaries

This article owns scholarly language interpretation within TEI P5 encoded texts: parallel passage correspondence, historical manuscripts, marginal notes and editorial apparatus. It does not duplicate the separate jobs of W3C Web Annotation anchoring, Wikibase lexical forms or SKOS multilingual labels. The reader should be able to distinguish the artifact, the transcription, the translation and a modern editor’s explanation at every significant decision.

For example, an apparently simple term in a damaged charter may appear with missing letters. If the translation supplies a confident noun but the source editor records uncertainty, the target-language reader has received a stronger claim than the original warrants. The right response is not to make the translation unpleasant; it is to preserve the limit of knowledge while explaining the most plausible reading.

First principles for evidence-preserving translation

1. Witness and translation are not the same object

A facsimile, diplomatic transcription, normalized edition and translation form different layers of evidence. Fluent modern writing must not replace what the source actually records. Test the principle with one actual segment from the edition. Ask what a source-language scholar can verify, what the translator has inferred and what a target-language reader may reasonably conclude. These answers should agree about the limits of evidence even when grammatical structures differ.

Keep a concise decision record: the source location, the relevant markup or image, the proposed target wording and the reason that wording is justified. A second reviewer should be able to reproduce the logic without simply trusting the translator’s confidence. When several readings remain possible, explicitly retain that interpretive range.

2. The header establishes provenance

The TEI header describes sources, titles, editions and responsibility. Translation requires an editorial decision about which claims belong to source description and which to the new edition. Test the principle with one actual segment from the edition. Ask what a source-language scholar can verify, what the translator has inferred and what a target-language reader may reasonably conclude. These answers should agree about the limits of evidence even when grammatical structures differ.

Keep a concise decision record: the source location, the relevant markup or image, the proposed target wording and the reason that wording is justified. A second reviewer should be able to reproduce the logic without simply trusting the translator’s confidence. When several readings remain possible, explicitly retain that interpretive range.

3. xml:lang describes real language

Language identification can appear on a passage or nested span. Use it for actual content language, not merely for where the file is stored. Test the principle with one actual segment from the edition. Ask what a source-language scholar can verify, what the translator has inferred and what a target-language reader may reasonably conclude. These answers should agree about the limits of evidence even when grammatical structures differ.

Keep a concise decision record: the source location, the relevant markup or image, the proposed target wording and the reason that wording is justified. A second reviewer should be able to reproduce the logic without simply trusting the translator’s confidence. When several readings remain possible, explicitly retain that interpretive range.

4. xml:id is a structural pointer

Identifiers and link targets make alignment possible. Their spelling is not a translation prompt. Test the principle with one actual segment from the edition. Ask what a source-language scholar can verify, what the translator has inferred and what a target-language reader may reasonably conclude. These answers should agree about the limits of evidence even when grammatical structures differ.

Keep a concise decision record: the source location, the relevant markup or image, the proposed target wording and the reason that wording is justified. A second reviewer should be able to reproduce the logic without simply trusting the translator’s confidence. When several readings remain possible, explicitly retain that interpretive range.

5. Alignment need not be one-to-one

Two source clauses may align to one target clause, and one source phrase may require several target phrases. Make the correspondence explicit rather than imposing a false sentence count. Test the principle with one actual segment from the edition. Ask what a source-language scholar can verify, what the translator has inferred and what a target-language reader may reasonably conclude. These answers should agree about the limits of evidence even when grammatical structures differ.

Keep a concise decision record: the source location, the relevant markup or image, the proposed target wording and the reason that wording is justified. A second reviewer should be able to reproduce the logic without simply trusting the translator’s confidence. When several readings remain possible, explicitly retain that interpretive range.

6. Uncertainty is a scholarly fact

Unclear writing, missing material and conjectural restoration must remain visible because they determine how confident the reader can be in an interpretation. Test the principle with one actual segment from the edition. Ask what a source-language scholar can verify, what the translator has inferred and what a target-language reader may reasonably conclude. These answers should agree about the limits of evidence even when grammatical structures differ.

Keep a concise decision record: the source location, the relevant markup or image, the proposed target wording and the reason that wording is justified. A second reviewer should be able to reproduce the logic without simply trusting the translator’s confidence. When several readings remain possible, explicitly retain that interpretive range.

7. Corrections are evidence

Original spellings, regularizations, scribal errors and editorial corrections show layers of textual history. Test the principle with one actual segment from the edition. Ask what a source-language scholar can verify, what the translator has inferred and what a target-language reader may reasonably conclude. These answers should agree about the limits of evidence even when grammatical structures differ.

Keep a concise decision record: the source location, the relevant markup or image, the proposed target wording and the reason that wording is justified. A second reviewer should be able to reproduce the logic without simply trusting the translator’s confidence. When several readings remain possible, explicitly retain that interpretive range.

8. Form matters

Poetry, speaker turns, line boundaries and marginal positions can be meaningful, not accidental formatting. Test the principle with one actual segment from the edition. Ask what a source-language scholar can verify, what the translator has inferred and what a target-language reader may reasonably conclude. These answers should agree about the limits of evidence even when grammatical structures differ.

Keep a concise decision record: the source location, the relevant markup or image, the proposed target wording and the reason that wording is justified. A second reviewer should be able to reproduce the logic without simply trusting the translator’s confidence. When several readings remain possible, explicitly retain that interpretive range.

9. Identity may be historically layered

People, places and dates can have variant names and calendar systems. Explain them with evidence rather than silently modernizing. Test the principle with one actual segment from the edition. Ask what a source-language scholar can verify, what the translator has inferred and what a target-language reader may reasonably conclude. These answers should agree about the limits of evidence even when grammatical structures differ.

Keep a concise decision record: the source location, the relevant markup or image, the proposed target wording and the reason that wording is justified. A second reviewer should be able to reproduce the logic without simply trusting the translator’s confidence. When several readings remain possible, explicitly retain that interpretive range.

10. Validation is dual

A TEI file can be structurally valid yet linguistically misleading, or linguistically faithful yet full of broken pointers. Both tests are required. Test the principle with one actual segment from the edition. Ask what a source-language scholar can verify, what the translator has inferred and what a target-language reader may reasonably conclude. These answers should agree about the limits of evidence even when grammatical structures differ.

Keep a concise decision record: the source location, the relevant markup or image, the proposed target wording and the reason that wording is justified. A second reviewer should be able to reproduce the logic without simply trusting the translator’s confidence. When several readings remain possible, explicitly retain that interpretive range.

A twelve-stage editorial and technical workflow

Use a short representative passage containing at least one identifier, one textual intervention and one alignment relation before translating a full collection. It makes the difference between textual accuracy and file integrity observable. A good project can roll back a mistaken interpretation without losing the original witness or silently changing references.

Stage 1: Identify the witness

Record manuscript identity, institution, facsimile and existing transcription. This is a verifiable action, not simply a task label. Capture the source evidence, target decision, reviewer and acceptance test for this stage. If an editor cannot tell whether a change alters the witness or only its explanation, pause that change and seek a source specialist’s judgment.

Ask two separate questions before progressing: does this target-language passage say no more than the source permits, and can the project resolve the same scholarly links after the change? The first is a philological question; the second is a technical one. Both matter because a good digital edition is an argument supported by inspectable evidence, not merely a nicely formatted reading copy.

Stage 2: Read TEI header

Examine sourceDesc, titleStmt, edition and responsibility. This is a verifiable action, not simply a task label. Capture the source evidence, target decision, reviewer and acceptance test for this stage. If an editor cannot tell whether a change alters the witness or only its explanation, pause that change and seek a source specialist’s judgment.

Ask two separate questions before progressing: does this target-language passage say no more than the source permits, and can the project resolve the same scholarly links after the change? The first is a philological question; the second is a technical one. Both matter because a good digital edition is an argument supported by inspectable evidence, not merely a nicely formatted reading copy.

Stage 3: Define editorial policy

Specify approaches for original spellings, uncertainty and supplied material. This is a verifiable action, not simply a task label. Capture the source evidence, target decision, reviewer and acceptance test for this stage. If an editor cannot tell whether a change alters the witness or only its explanation, pause that change and seek a source specialist’s judgment.

Ask two separate questions before progressing: does this target-language passage say no more than the source permits, and can the project resolve the same scholarly links after the change? The first is a philological question; the second is a technical one. Both matter because a good digital edition is an argument supported by inspectable evidence, not merely a nicely formatted reading copy.

Stage 4: Protect keys

Freeze xml:id and pointer targets pending any full migration plan. This is a verifiable action, not simply a task label. Capture the source evidence, target decision, reviewer and acceptance test for this stage. If an editor cannot tell whether a change alters the witness or only its explanation, pause that change and seek a source specialist’s judgment.

Ask two separate questions before progressing: does this target-language passage say no more than the source permits, and can the project resolve the same scholarly links after the change? The first is a philological question; the second is a technical one. Both matter because a good digital edition is an argument supported by inspectable evidence, not merely a nicely formatted reading copy.

Stage 5: Segment meaning

Choose passage boundaries that reflect interpretation, not merely punctuation. This is a verifiable action, not simply a task label. Capture the source evidence, target decision, reviewer and acceptance test for this stage. If an editor cannot tell whether a change alters the witness or only its explanation, pause that change and seek a source specialist’s judgment.

Ask two separate questions before progressing: does this target-language passage say no more than the source permits, and can the project resolve the same scholarly links after the change? The first is a philological question; the second is a technical one. Both matter because a good digital edition is an argument supported by inspectable evidence, not merely a nicely formatted reading copy.

Stage 6: Translate responsibly

Preserve claims, voices, limits and ambiguity. This is a verifiable action, not simply a task label. Capture the source evidence, target decision, reviewer and acceptance test for this stage. If an editor cannot tell whether a change alters the witness or only its explanation, pause that change and seek a source specialist’s judgment.

Ask two separate questions before progressing: does this target-language passage say no more than the source permits, and can the project resolve the same scholarly links after the change? The first is a philological question; the second is a technical one. Both matter because a good digital edition is an argument supported by inspectable evidence, not merely a nicely formatted reading copy.

Stage 7: Set language metadata

Identify language of translated passages and foreign quotations. This is a verifiable action, not simply a task label. Capture the source evidence, target decision, reviewer and acceptance test for this stage. If an editor cannot tell whether a change alters the witness or only its explanation, pause that change and seek a source specialist’s judgment.

Ask two separate questions before progressing: does this target-language passage say no more than the source permits, and can the project resolve the same scholarly links after the change? The first is a philological question; the second is a technical one. Both matter because a good digital edition is an argument supported by inspectable evidence, not merely a nicely formatted reading copy.

Stage 8: Review alignment

Represent justified many-to-one and one-to-many correspondences. This is a verifiable action, not simply a task label. Capture the source evidence, target decision, reviewer and acceptance test for this stage. If an editor cannot tell whether a change alters the witness or only its explanation, pause that change and seek a source specialist’s judgment.

Ask two separate questions before progressing: does this target-language passage say no more than the source permits, and can the project resolve the same scholarly links after the change? The first is a philological question; the second is a technical one. Both matter because a good digital edition is an argument supported by inspectable evidence, not merely a nicely formatted reading copy.

Stage 9: Review interventions

Track additions, deletions, unclear text and corrections. This is a verifiable action, not simply a task label. Capture the source evidence, target decision, reviewer and acceptance test for this stage. If an editor cannot tell whether a change alters the witness or only its explanation, pause that change and seek a source specialist’s judgment.

Ask two separate questions before progressing: does this target-language passage say no more than the source permits, and can the project resolve the same scholarly links after the change? The first is a philological question; the second is a technical one. Both matter because a good digital edition is an argument supported by inspectable evidence, not merely a nicely formatted reading copy.

Stage 10: Validate document

Check project schema, unique IDs and resolved pointers. This is a verifiable action, not simply a task label. Capture the source evidence, target decision, reviewer and acceptance test for this stage. If an editor cannot tell whether a change alters the witness or only its explanation, pause that change and seek a source specialist’s judgment.

Ask two separate questions before progressing: does this target-language passage say no more than the source permits, and can the project resolve the same scholarly links after the change? The first is a philological question; the second is a technical one. Both matter because a good digital edition is an argument supported by inspectable evidence, not merely a nicely formatted reading copy.

Stage 11: Compare facsimile

Read representative difficult portions against actual source evidence. This is a verifiable action, not simply a task label. Capture the source evidence, target decision, reviewer and acceptance test for this stage. If an editor cannot tell whether a change alters the witness or only its explanation, pause that change and seek a source specialist’s judgment.

Ask two separate questions before progressing: does this target-language passage say no more than the source permits, and can the project resolve the same scholarly links after the change? The first is a philological question; the second is a technical one. Both matter because a good digital edition is an argument supported by inspectable evidence, not merely a nicely formatted reading copy.

Stage 12: Release apparatus

Document witnesses, editor responsibility, changes and unresolved readings. This is a verifiable action, not simply a task label. Capture the source evidence, target decision, reviewer and acceptance test for this stage. If an editor cannot tell whether a change alters the witness or only its explanation, pause that change and seek a source specialist’s judgment.

Ask two separate questions before progressing: does this target-language passage say no more than the source permits, and can the project resolve the same scholarly links after the change? The first is a philological question; the second is a technical one. Both matter because a good digital edition is an argument supported by inspectable evidence, not merely a nicely formatted reading copy.

Twenty-six worked scholarly translation cases

Worked case 1: Header title

Witness or encoding situation: A TEI header includes the manuscript’s source title. Common wrong move: Replacing source title with an unsignaled modern translation. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Keep original title and distinguish translation. Why: An edition’s bibliographic identity is separate from a helpful display name. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Check titleStmt and sourceDesc. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 2: Shelfmark

Witness or encoding situation: An archive assigns a stable manuscript shelfmark. Common wrong move: Translating or simplifying the shelfmark. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Preserve exact catalog identifier. Why: Researchers must locate the same artifact. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Search institutional catalog. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 3: Latin quotation

Witness or encoding situation: An English translation retains a Latin phrase. Common wrong move: Using English language markup on the Latin span. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Mark actual language on the span. Why: Language metadata is content description, not decoration. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Inspect xml:lang inheritance. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 4: Source IDs

Witness or encoding situation: Source and target segments use src-17 and trg-17. Common wrong move: Renaming one ID without updating pointers. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Keep IDs or migrate every dependent link. Why: A valid looking translation can have broken alignment. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Resolve pointer targets. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 5: linkGrp target

Witness or encoding situation: An alignment group connects two passages. Common wrong move: The target references a nonexistent ID. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Repair reference to actual segment. Why: Scholarly alignment requires resolvable relationships. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Validate references. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 6: Many-to-one alignment

Witness or encoding situation: Two clauses in one language become one target sentence. Common wrong move: Forcing inaccurate one-to-one equivalence. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Represent two-to-one alignment when justified. Why: Syntax varies across languages. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Compare the full span. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 7: One-to-many expansion

Witness or encoding situation: A compact source term requires two target clauses. Common wrong move: Adding an unsupported independent claim. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Align explanatory target material transparently. Why: Translation and editorial clarification have different status. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Check source claim. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 8: Illegible word

Witness or encoding situation: A damaged word is uncertain. Common wrong move: Writing one confident reconstructed noun. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Keep uncertainty visible and label conjecture. Why: A guess must not become observed evidence. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Inspect facsimile. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 9: Missing text

Witness or encoding situation: A physical hole causes a lacuna. Common wrong move: Inventing a fluent replacement sentence. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Mark the gap and any documented restoration. Why: Manuscript absence is part of the evidence. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Compare damaged folio. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 10: Supplied letters

Witness or encoding situation: An editor adds letters at torn edge. Common wrong move: Showing supply as witnessed text. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Preserve editorial supply distinction. Why: Readers should know which letters are reconstructed. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Inspect apparatus. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 11: Historical spelling

Witness or encoding situation: A medieval word has nonmodern spelling. Common wrong move: Silently normalizing all witness forms. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Preserve a policy for orig and reg. Why: Spelling may indicate dialect or dating. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Compare variants. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 12: Scribal correction

Witness or encoding situation: An error has a proposed corrected reading. Common wrong move: Suppressing the original reading. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Show correction under transparent policy. Why: Critical editing needs visible alternatives. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Inspect sic and corr. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 13: Marginal addition

Witness or encoding situation: Another hand adds words in the margin. Common wrong move: Inserting them into main sentence without note. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Keep relation to addition location. Why: Later material can alter the source’s history. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Check add location. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 14: Deleted phrase

Witness or encoding situation: Words are struck through. Common wrong move: Restoring them as active original text. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Retain deletion marking. Why: The artifact shows a revision. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Check del markup. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 15: Hand change

Witness or encoding situation: A manuscript changes scribal hands. Common wrong move: Crediting all writing to the first scribe. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Record hand identity cautiously. Why: Attribution is evidence-dependent. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Check handShift and image. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 16: Page boundary

Witness or encoding situation: An argument crosses a folio break. Common wrong move: Inventing sentence boundaries at every page. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Keep page reference without altering syntax. Why: Page navigation and sentence structure differ. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Check pb/facs. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 17: Facsimile link

Witness or encoding situation: A line points to a digitized folio. Common wrong move: Translating the image URI. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Preserve working facsimile reference. Why: Readings should be testable against source images. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Open image link. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 18: Historical person

Witness or encoding situation: A name has multiple historical forms. Common wrong move: Imposing one present-day form without note. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Retain witness form and give verified identity note. Why: Original naming is historically significant. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Check prosopography. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 19: Historical town

Witness or encoding situation: A town is named under an older polity. Common wrong move: Replacing with a modern country label unqualified. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Preserve source form and annotate identification. Why: Anachronism can distort place meaning. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Consult historical gazetteer. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 20: Regnal year

Witness or encoding situation: A charter dates an event by a ruler’s reign. Common wrong move: Inserting a modern date without conversion method. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Keep original date and explain conversion evidence. Why: Historical calendar conversions can be uncertain. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Review dating apparatus. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 21: Verse line

Witness or encoding situation: Poetry uses line and stanza structure. Common wrong move: Flattening all lines into paragraphs. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Preserve meaningful poetic organization or document editorial choice. Why: Form may influence meaning. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Compare line groups. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 22: Speaker

Witness or encoding situation: A drama changes speakers. Common wrong move: Assigning translated phrase to wrong person. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Keep speaker divisions and names. Why: Voice attribution changes interpretation. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Check each speech boundary. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 23: Bibliographic note

Witness or encoding situation: A note references a historic edition. Common wrong move: Inventing a new official title. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Retain locatable citation with optional gloss. Why: Readers need to inspect the exact cited work. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Search bibliographic record. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 24: Editorial annotation

Witness or encoding situation: A note explains uncertain word choice. Common wrong move: Merging note into original sentence. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Keep text distinct from commentary. Why: Readers must distinguish translation and editor argument. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Compare note display. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 25: Stand-off analysis

Witness or encoding situation: Separate annotation points to text segments. Common wrong move: Deleting links that are not visible. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Preserve stand-off relationships. Why: Corpus processing may depend on them. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Run pointer test. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Worked case 26: Responsibility

Witness or encoding situation: Several transcribers and translators contributed. Common wrong move: Attributing every choice to one translator. Before changing anything, distinguish observed textual evidence from an editor’s reconstruction or a machine reference.

Correct editorial treatment: Record roles in responsible statements. Why: Scholarly accountability matters. Explain to the target-language reader what is attested, what is interpreted and what uncertainty remains. A stylistic preference cannot itself settle missing manuscript evidence or make a disconnected XML pointer meaningful.

Independent verification: Check respStmt. Ideally, a second reviewer should inspect the manuscript, transcription or TEI representation and then test the translated passage. Where image access is unavailable, flag the evidence limit. Where a pointer is structural, use a schema and link check rather than relying on visual appearance.

Transfer the lesson to another manuscript by first naming the historical claim at stake, then identifying its source layer and finally deciding what target-language explanation preserves that layer. This prevents a frequent scholarly error: allowing a translator’s fluent prose to conceal a chain of speculative choices that a reader should be able to challenge.

Failure modes: systematic diagnosis

Failure 1: False certainty

Unclear letters become a confident quotation. Repair strategy: Preserve uncertainty. Document which source passage, reference or witness layer was affected and how the corrected target edition was tested. A green XML validator is not enough to establish good historical interpretation, and a good paragraph is not enough to establish functioning links. Retest both where relevant.

Failure 2: Orphaned pointers

Target references disappear. Repair strategy: Validate ID and URI links. Document which source passage, reference or witness layer was affected and how the corrected target edition was tested. A green XML validator is not enough to establish good historical interpretation, and a good paragraph is not enough to establish functioning links. Retest both where relevant.

Failure 3: Witness overwritten

Only modern smooth text remains. Repair strategy: Keep transcription and translation distinct. Document which source passage, reference or witness layer was affected and how the corrected target edition was tested. A green XML validator is not enough to establish good historical interpretation, and a good paragraph is not enough to establish functioning links. Retest both where relevant.

Failure 4: Invisible editing

Supplied or corrected wording is unmarked. Repair strategy: Reopen apparatus and evidence. Document which source passage, reference or witness layer was affected and how the corrected target edition was tested. A green XML validator is not enough to establish good historical interpretation, and a good paragraph is not enough to establish functioning links. Retest both where relevant.

Failure 5: Forced segmentation

Every source line gets exactly one target line regardless of syntax. Repair strategy: Allow justified non-one-to-one links. Document which source passage, reference or witness layer was affected and how the corrected target edition was tested. A green XML validator is not enough to establish good historical interpretation, and a good paragraph is not enough to establish functioning links. Retest both where relevant.

Failure 6: Language bleed

Foreign passages inherit wrong language. Repair strategy: Set xml:lang on relevant spans. Document which source passage, reference or witness layer was affected and how the corrected target edition was tested. A green XML validator is not enough to establish good historical interpretation, and a good paragraph is not enough to establish functioning links. Retest both where relevant.

Failure 7: Image route failure

Page link opens wrong folio. Repair strategy: Check facsimile targets. Document which source passage, reference or witness layer was affected and how the corrected target edition was tested. A green XML validator is not enough to establish good historical interpretation, and a good paragraph is not enough to establish functioning links. Retest both where relevant.

Failure 8: Misattributed annotation

Marginal note becomes main author’s claim. Repair strategy: Preserve hand and note roles. Document which source passage, reference or witness layer was affected and how the corrected target edition was tested. A green XML validator is not enough to establish good historical interpretation, and a good paragraph is not enough to establish functioning links. Retest both where relevant.

Failure 9: Historical anachronism

Old place name silently becomes modern entity. Repair strategy: Document name interpretation. Document which source passage, reference or witness layer was affected and how the corrected target edition was tested. A green XML validator is not enough to establish good historical interpretation, and a good paragraph is not enough to establish functioning links. Retest both where relevant.

Failure 10: Invalid TEI

Reader view looks fine but schema is broken. Repair strategy: Run TEI validation. Document which source passage, reference or witness layer was affected and how the corrected target edition was tested. A green XML validator is not enough to establish good historical interpretation, and a good paragraph is not enough to establish functioning links. Retest both where relevant.

Practice and transfer with model reasoning

Practice 1: Header title

A scholarly team gives you this case: A TEI header includes the manuscript’s source title. An initial translator suggests Replacing source title with an unsignaled modern translation. Decide whether the issue concerns evidence, interpretation, alignment or identification. Draft a better treatment and identify the exact check that would make a false reading visible.

Model reasoning: Keep original title and distinguish translation, because An edition’s bibliographic identity is separate from a helpful display name. Perform the check: Check titleStmt and sourceDesc. A complete answer distinguishes source from editorial intervention, rather than merely improving the sentence’s fluency. The criterion is whether another scholar can reconstruct why the target passage was justified.

Practice 2: Shelfmark

A scholarly team gives you this case: An archive assigns a stable manuscript shelfmark. An initial translator suggests Translating or simplifying the shelfmark. Decide whether the issue concerns evidence, interpretation, alignment or identification. Draft a better treatment and identify the exact check that would make a false reading visible.

Model reasoning: Preserve exact catalog identifier, because Researchers must locate the same artifact. Perform the check: Search institutional catalog. A complete answer distinguishes source from editorial intervention, rather than merely improving the sentence’s fluency. The criterion is whether another scholar can reconstruct why the target passage was justified.

Practice 3: Latin quotation

A scholarly team gives you this case: An English translation retains a Latin phrase. An initial translator suggests Using English language markup on the Latin span. Decide whether the issue concerns evidence, interpretation, alignment or identification. Draft a better treatment and identify the exact check that would make a false reading visible.

Model reasoning: Mark actual language on the span, because Language metadata is content description, not decoration. Perform the check: Inspect xml:lang inheritance. A complete answer distinguishes source from editorial intervention, rather than merely improving the sentence’s fluency. The criterion is whether another scholar can reconstruct why the target passage was justified.

Practice 4: Source IDs

A scholarly team gives you this case: Source and target segments use src-17 and trg-17. An initial translator suggests Renaming one ID without updating pointers. Decide whether the issue concerns evidence, interpretation, alignment or identification. Draft a better treatment and identify the exact check that would make a false reading visible.

Model reasoning: Keep IDs or migrate every dependent link, because A valid looking translation can have broken alignment. Perform the check: Resolve pointer targets. A complete answer distinguishes source from editorial intervention, rather than merely improving the sentence’s fluency. The criterion is whether another scholar can reconstruct why the target passage was justified.

Practice 5: linkGrp target

A scholarly team gives you this case: An alignment group connects two passages. An initial translator suggests The target references a nonexistent ID. Decide whether the issue concerns evidence, interpretation, alignment or identification. Draft a better treatment and identify the exact check that would make a false reading visible.

Model reasoning: Repair reference to actual segment, because Scholarly alignment requires resolvable relationships. Perform the check: Validate references. A complete answer distinguishes source from editorial intervention, rather than merely improving the sentence’s fluency. The criterion is whether another scholar can reconstruct why the target passage was justified.

Practice 6: Many-to-one alignment

A scholarly team gives you this case: Two clauses in one language become one target sentence. An initial translator suggests Forcing inaccurate one-to-one equivalence. Decide whether the issue concerns evidence, interpretation, alignment or identification. Draft a better treatment and identify the exact check that would make a false reading visible.

Model reasoning: Represent two-to-one alignment when justified, because Syntax varies across languages. Perform the check: Compare the full span. A complete answer distinguishes source from editorial intervention, rather than merely improving the sentence’s fluency. The criterion is whether another scholar can reconstruct why the target passage was justified.

Practice 7: One-to-many expansion

A scholarly team gives you this case: A compact source term requires two target clauses. An initial translator suggests Adding an unsupported independent claim. Decide whether the issue concerns evidence, interpretation, alignment or identification. Draft a better treatment and identify the exact check that would make a false reading visible.

Model reasoning: Align explanatory target material transparently, because Translation and editorial clarification have different status. Perform the check: Check source claim. A complete answer distinguishes source from editorial intervention, rather than merely improving the sentence’s fluency. The criterion is whether another scholar can reconstruct why the target passage was justified.

Practice 8: Illegible word

A scholarly team gives you this case: A damaged word is uncertain. An initial translator suggests Writing one confident reconstructed noun. Decide whether the issue concerns evidence, interpretation, alignment or identification. Draft a better treatment and identify the exact check that would make a false reading visible.

Model reasoning: Keep uncertainty visible and label conjecture, because A guess must not become observed evidence. Perform the check: Inspect facsimile. A complete answer distinguishes source from editorial intervention, rather than merely improving the sentence’s fluency. The criterion is whether another scholar can reconstruct why the target passage was justified.

Practice 9: Missing text

A scholarly team gives you this case: A physical hole causes a lacuna. An initial translator suggests Inventing a fluent replacement sentence. Decide whether the issue concerns evidence, interpretation, alignment or identification. Draft a better treatment and identify the exact check that would make a false reading visible.

Model reasoning: Mark the gap and any documented restoration, because Manuscript absence is part of the evidence. Perform the check: Compare damaged folio. A complete answer distinguishes source from editorial intervention, rather than merely improving the sentence’s fluency. The criterion is whether another scholar can reconstruct why the target passage was justified.

Practice 10: Supplied letters

A scholarly team gives you this case: An editor adds letters at torn edge. An initial translator suggests Showing supply as witnessed text. Decide whether the issue concerns evidence, interpretation, alignment or identification. Draft a better treatment and identify the exact check that would make a false reading visible.

Model reasoning: Preserve editorial supply distinction, because Readers should know which letters are reconstructed. Perform the check: Inspect apparatus. A complete answer distinguishes source from editorial intervention, rather than merely improving the sentence’s fluency. The criterion is whether another scholar can reconstruct why the target passage was justified.

Practice 11: Historical spelling

A scholarly team gives you this case: A medieval word has nonmodern spelling. An initial translator suggests Silently normalizing all witness forms. Decide whether the issue concerns evidence, interpretation, alignment or identification. Draft a better treatment and identify the exact check that would make a false reading visible.

Model reasoning: Preserve a policy for orig and reg, because Spelling may indicate dialect or dating. Perform the check: Compare variants. A complete answer distinguishes source from editorial intervention, rather than merely improving the sentence’s fluency. The criterion is whether another scholar can reconstruct why the target passage was justified.

Practice 12: Scribal correction

A scholarly team gives you this case: An error has a proposed corrected reading. An initial translator suggests Suppressing the original reading. Decide whether the issue concerns evidence, interpretation, alignment or identification. Draft a better treatment and identify the exact check that would make a false reading visible.

Model reasoning: Show correction under transparent policy, because Critical editing needs visible alternatives. Perform the check: Inspect sic and corr. A complete answer distinguishes source from editorial intervention, rather than merely improving the sentence’s fluency. The criterion is whether another scholar can reconstruct why the target passage was justified.

Frequently asked questions

Does TEI prescribe one translation alignment method?

TEI supports several linking and segmentation mechanisms; choose one that fits the edition and document it. Apply the answer under the edition’s published editorial policy. TEI provides a vocabulary for recording evidence but cannot make unsupported historical guesses true. When uncertainty persists, document it rather than silently selecting the most attractive interpretation.

Should xml:id be translated?

No, it is an identifier, not reader-facing prose. Apply the answer under the edition’s published editorial policy. TEI provides a vocabulary for recording evidence but cannot make unsupported historical guesses true. When uncertainty persists, document it rather than silently selecting the most attractive interpretation.

Can alignment be many-to-one?

Yes, correspondence should follow actual interpretive units rather than forced sentence equality. Apply the answer under the edition’s published editorial policy. TEI provides a vocabulary for recording evidence but cannot make unsupported historical guesses true. When uncertainty persists, document it rather than silently selecting the most attractive interpretation.

What does xml:lang establish?

It identifies the language of the marked content, often supporting tools and analysis. Apply the answer under the edition’s published editorial policy. TEI provides a vocabulary for recording evidence but cannot make unsupported historical guesses true. When uncertainty persists, document it rather than silently selecting the most attractive interpretation.

How do I render an illegible word?

Preserve the uncertainty and identify any proposed reconstruction as editorial. Apply the answer under the edition’s published editorial policy. TEI provides a vocabulary for recording evidence but cannot make unsupported historical guesses true. When uncertainty persists, document it rather than silently selecting the most attractive interpretation.

Why retain a diplomatic transcription?

It preserves evidence of the witness that polished translation may conceal. Apply the answer under the edition’s published editorial policy. TEI provides a vocabulary for recording evidence but cannot make unsupported historical guesses true. When uncertainty persists, document it rather than silently selecting the most attractive interpretation.

Should historic names be modernized?

Usually preserve witnessed forms and provide sourced modern identification as a note. Apply the answer under the edition’s published editorial policy. TEI provides a vocabulary for recording evidence but cannot make unsupported historical guesses true. When uncertainty persists, document it rather than silently selecting the most attractive interpretation.

May translators change source markup?

Only with an approved editorial policy and reference validation. Apply the answer under the edition’s published editorial policy. TEI provides a vocabulary for recording evidence but cannot make unsupported historical guesses true. When uncertainty persists, document it rather than silently selecting the most attractive interpretation.

Why keep the TEI header?

It captures source identity, editorial responsibility and encoding decisions. Apply the answer under the edition’s published editorial policy. TEI provides a vocabulary for recording evidence but cannot make unsupported historical guesses true. When uncertainty persists, document it rather than silently selecting the most attractive interpretation.

Does XML validation prove accuracy?

No, validation tests structure, not historical interpretation or linguistic faithfulness. Apply the answer under the edition’s published editorial policy. TEI provides a vocabulary for recording evidence but cannot make unsupported historical guesses true. When uncertainty persists, document it rather than silently selecting the most attractive interpretation.

How can image references be tested?

Open them and compare each cited page or region against the witness. Apply the answer under the edition’s published editorial policy. TEI provides a vocabulary for recording evidence but cannot make unsupported historical guesses true. When uncertainty persists, document it rather than silently selecting the most attractive interpretation.

Who should review?

A target-language expert, source-language subject specialist and TEI editor. Apply the answer under the edition’s published editorial policy. TEI provides a vocabulary for recording evidence but cannot make unsupported historical guesses true. When uncertainty persists, document it rather than silently selecting the most attractive interpretation.

Scholarly release checklist

  1. Verify manuscript witness and source bibliographic identity.
  2. Keep header responsibility and edition claims accurate.
  3. Check xml:lang in source and target passages.
  4. Confirm every referenced xml:id and target.
  5. Document non-one-to-one passage alignment.
  6. Inspect additions, deletions, supplied readings and gaps.
  7. Review lines, speakers and facsimile links.
  8. Validate XML against the project TEI schema.
  9. Ask a source specialist to review uncertain interpretations.
  10. Publish editorial policy and limitations.

Primary sources and contextual next routes

The technical source is the TEI P5 Guidelines, Linking, Segmentation, and Alignment, which identifies version 4.12.0 as updated 28 July 2026. Editorial discoverability should follow Google Search Central people-first content guidance, with clear distinctions between original evidence and explanatory teaching.

For language foundations, visit the Vocabulary Learning Hub and How English Works. Connected but distinct translation guides include Web Annotation selectors, Wikibase Lexemes, SKOS concept labels and IIIF museum language maps. They own different technical jobs.

The trustworthy result is not simply a convincing modern sentence. It is a target-language interpretation whose links, original evidence, editorial interventions and limitations remain inspectable by someone who disagrees with the translator.