Export, reuse & cite

A digital manuscript project is only as durable as the formats it commits to. An Archetype instance is built so that the scholarly work it holds — image regions, transcriptions, palaeographic classifications, and the links between them — can leave the platform in the standard vocabularies that libraries, viewers, and other research tools already understand. Nothing is trapped in a proprietary schema. The same data that powers search and browse inside the platform is served, on demand, as W3C Web Annotations, IIIF resources, and TEI documents.

Interoperability here is a derived view over the canonical data model, not a second copy. An image region is a Graph; a text-to-region link is a reference embedded in the transcription markup. The standards endpoints translate those canonical objects into open formats at read time, so an export is always faithful to the live corpus rather than a stale dump.

Quick exports for your own research

Most researchers need only these, all available from the public site without an account:

• Search results download as CSV, JSON, or BibTeX (for manuscripts, scribes and hands), either the current page or the whole result set. See Search & browse.
• A manuscript’s glyph annotations download as CSV from its annotation gallery.
• A transcription or translation downloads as a standalone TEI P5 document, and a published manuscript description as TEI too.
• A whole manuscript opens in any IIIF viewer, such as Mirador, from its manifest URL.
• A lightbox arrangement exports as PDF, images, JSON or TEI. See Lightbox & worksets.

To cite what you used, see Citing Archetype. The rest of this page describes the standard formats in detail, for anyone connecting another tool.

W3C Web Annotations (JSON-LD)

The annotation model — glyphs boxed on a page image, classified against an allograph taxonomy, and linked to points in a transcription — is exposed as W3C Web Annotations, the standard the whole annotation ecosystem shares.

• A single region is served as an Annotation whose target carries both a FragmentSelector (a xywh= bounding box) and an SvgSelector (the exact polygon), with a motivation chosen by annotation type — describing for a palaeographic glyph, identifying for a text anchor, commenting for an editorial note. Classifications appear as a SpecificResource body pointing back to the allograph.
• A whole transcription is served as an AnnotationPage, one annotation per linked element, combining a TextQuoteSelector (the exact phrase in the text) with the image region it corresponds to.

These endpoints are public and read-only, so external scholarly tools can consume a corpus’s annotations directly over HTTP. One deliberate detail worth noting for reuse: stored geometry uses a legacy bottom-left origin, and the converters flip it into the top-left origin that IIIF and Media Fragments expect, so selectors land in the right place in any standard viewer.

IIIF: manifests and search-within

Every manuscript part can be requested as a IIIF Presentation 3.0 Manifest — one Canvas per image, a painting annotation referencing the image on the IIIF image server, and, where a transcription exists, a supplementary AnnotationPage whose text is anchored to the correct image regions, plus a page of the image’s glyph annotations. This is what lets a page from your corpus open in Mirador, Universal Viewer, or any IIIF-compliant client, side by side with material from other institutions.

Each manifest also advertises a IIIF Content Search 2.0 service, so a viewer can search within a manuscript and have matching passages boxed on the page image.

Search-within resolves to transcription phrases an editor has manually linked to a region. Prose that carries no region link has no coordinates to return — arbitrary word-level search would require word coordinates the platform does not yet store. This is a documented boundary of the feature, not a defect.

TEI import and export

Transcriptions are stored as TEI P5, the scholarly standard for encoded text. Any transcription can be downloaded as a standalone, titled TEI document, and an existing TEI file can be imported from the backoffice as a new draft. A manuscript’s published msDesc description downloads the same way, as a TEI document with the description in its teiHeader. Older DigiPal-style markup (for example, in a legacy backup) is converted to TEI by a round-trip-verified command: a row is only converted if it converts back to its original form byte-for-canonical-byte, and rows that fail are left untouched and reported rather than silently mangled. Well-formedness validation and formatting endpoints are also exposed for editing tools.

HTR import (PAGE-XML / ALTO)

Output from handwritten-text-recognition engines can be imported directly through the management API (there is no backoffice screen for it yet). The platform parses both PAGE-XML and ALTO — the two formats HTR tools such as Transkribus and eScriptorium emit — into a draft transcription. Each recognised line becomes a TEI <seg type="line">, and where the engine supplies a line polygon, the platform materialises a region and wires the corresp link automatically, so an HTR pass arrives already connected to the page image. See annotations for how those regions then behave in the editor.