TEI 拡張語彙 / Soranoha TEI extension vocabulary

Namespace IRI: https://w3id.org/soranoha/ns/tei Conventional prefix: snh: Current vocabulary version: 1

Published TEI documents bind this namespace IRI on the root element.

Soranoha publishes TEI P5. Where the source markup carries information TEI has no element for, that information is recorded as an attribute in this namespace rather than by inventing an element. Every published TEI root binds the namespace and declares the vocabulary version:

<TEI xmlns="http://www.tei-c.org/ns/1.0"
     xmlns:snh="https://w3id.org/soranoha/ns/tei"
     snh:vocab-version="1">

These attributes are derived audit projections. They record what the parser established about the Aozora Bunko source; they do not replace the source records from which they were derived, and a consumer may ignore them without losing the transcription.

Versioning

The namespace IRI identifies the extension and never changes. The vocabulary's major version is not implied by that IRI, so snh:vocab-version carries it separately. A breaking vocabulary change increments this version, whereas an additive attribute may keep the same major version when compatibility is documented.

The profile hashes recorded in a work's tei-validation.json (odd_hash, rng_hash, schematron_hash) answer a different question. They change on any profile edit, additive ones included, so they say which profile validated a document but cannot tell you whether an existing snh:* reader is still correct. That is what the vocabulary version is for.

Version 1 contains attributes only. Custom elements in this namespace are not part of it.

Attributes

All are optional and may appear on any element, subject to the constraints below. Only the first three are emitted by the current pipeline; the rest are declared for projections that are specified but not yet produced.

Attribute	Value	Meaning
`snh:vocab-version`	`1`	Vocabulary major version. Required on the root of any document carrying `snh:*` attributes.
`snh:layout-kind`	token, or space-separated tokens when scopes co-apply	Layout fact projected verbatim from the parser IR. Paragraph layouts are `burasage`, `chitsuki`, `jisage`, `jizume`, `line-jisage`; inline scopes are `baseline-position`, `chitsuki`, `emphasis`, `exponent`, `font-size`, `fraction`, `keigakomi`, `small-script`, `tcy`, `yokogumi`.
`snh:layout-params`	`key=value` pairs joined by `;`	Parameters of the layout fact, for example `indent=2` or `align=right;offset-from-end=1`.
`snh:preservation-record`	`r` followed by six digits	Record id in the publication-preservation sidecar.
`snh:ir-pointer`	JSON Pointer	Location in the parser IR of the projected fact.
`snh:source-pointer`	pointer string	Location in the source or AAT of the projected fact.
`snh:source-marker`	string	Original Aozora Bunko marker, kept when a TEI-native rendering normalizes it away.
`snh:accent-code`	string	Aozora Bunko accent notation code.
`snh:style-source`	string	Aozora Bunko style source marker.
`snh:figure-class`	string	Aozora Bunko figure class.

Permitted values of snh:vocab-version are declared in the ODD and enforced by the generated Relax NG schema. The initial version is 1.

Source spans

Most elements in a published body carry @source, pointing at a note of type source-span in the back matter. Those notes are the offset table: they say which run of the source text an element came from.

A note is empty, because everything about the extent is in its reference:

<note type="source-span" xml:id="source-694-733" n="20"/>

xml:id has the form source-START-END. START and END are offsets in UTF-8 bytes into the decoded primary text, which sourceDesc identifies by its idno of type primary-text-hash. Decode the source archive's text member, encode it as UTF-8, and slice.
n is the line the extent begins on. It is absent when the source did not record one, rather than carrying a placeholder.

The header declares this unit in a refsDecl pointing back at this document.

Notes with xml:id prefixed by parser-source belong to parser diagnostics rather than body elements; they carry their byte extent as JSON in element content.

Schematron rule identifiers

Constraint violations are published in each work's tei-validation.json as findings[].rule_id. The identifiers are stable names, not sequence numbers. Every identifier below comes from the Schematron layer; a Relax NG failure is reported with the constant rule_id relax-ng instead.

Rule ID	Severity	Requires
`snh-tei-header-title`	error	A main title in `teiHeader/fileDesc/titleStmt` (a `title` that is not one of the subordinate forms `sub`, `sub-reading`, `reading`, or `original`).
`snh-tei-header-source-work-id`	error	An Aozora Bunko work ID or source work identifier in the header.
`snh-header-language-declared`	error	At least one `profileDesc/langUsage/language` with a non-empty `@ident`.
`snh-publication-licence`	error	A rights grant in `publicationStmt/availability/licence` with a non-empty `@target`.
`snh-ruby-complete`	error	Ruby structures contain both base and reading components.
`snh-ruby-base-non-empty`	error	A ruby base contains text or a declared character.
`snh-ruby-reading-non-empty`	error	A ruby reading is not empty.
`snh-gaiji-reference`	error	Gaiji retains its original marker and either a resolved replacement or unresolved status.
`snh-gaiji-chardecl-resolution`	error	A gaiji `@ref` fragment resolves to a `charDecl/char` in the same document.
`snh-char-resolution-form`	error	A `charDecl/char` declares at least one of `mapping`, `unicodeProp`, `localProp`, `desc`.
`snh-figure-accessibility`	warning	A figure carries a nonempty description or caption, so the image it names has alternative text. Most Aozora Bunko illustration markers supply a filename and pixel dimensions and no description, so this fires on the source's own form rather than on anything the conversion dropped.
`snh-source-span-reference`	error	Elements claiming source spans point to valid span identifiers.
`snh-source-span-target-exists`	error	Each fragment id in `@source` resolves to an `@xml:id` in the same document.
`snh-transcription-vs-annotation`	warning	No element of TEI's analysis module appears inside `tei:text` unless the header declares it. Stand-off anchors and a `standOff` sibling are outside the rule, which is what the supported enrichment path writes.
`snh-vocab-version-declared`	error	A root carrying `snh:*` attributes declares `snh:vocab-version`.
`snh-preservation-record-shape`	error	`snh:preservation-record` values have the form `r000000`.
`snh-layout-params-shape`	error	`snh:layout-params` uses the `key=value` grammar above.

Rights in the header

Every published work states its own terms in fileDesc/publicationStmt/availability, with one licence element per rights layer (one for the standing of the underlying work, which is public domain or a Creative Commons Attribution licence, and one for Soranoha's CC0 grant over the encoding, which also carries a ptr to the full statement). The encoding grant is rendered from the one publication policy the release manifest hashes and the work's standing from the same value the manifest records in works[].rights, so a detached TEI file and the signed release record cannot state different terms. snh-publication-licence makes their absence a validation failure. See rights and licensing.

A schema pass does not establish source fidelity or publication rights. See TEI validation for how the two validation layers run, and source accountability for the separate fidelity evidence.

Provenance of this document

The vocabulary is defined by soranoha/schemas/tei-profile.odd, from which the Relax NG and Schematron are generated. Where this page and the ODD disagree, the ODD is authoritative.