Skip to main content

Conceptual Document Model

Purpose​

The document model is the conceptual core of the Open Manuscript Initiative (OMI).

Unlike traditional word processors, OMI does not treat a manuscript as formatted text. Instead, a manuscript is represented as a structured scholarly object composed of semantic entities and their relationships.

Presentation (PDF, HTML, JATS XML, EPUB, DOCX, etc.) is always generated from this semantic representation.

Design Principles

The document model is built upon several principles.

Semantic First​

Every element has meaning before appearance.

Formatting is a consequence of semantics, not the other way around.

Separation of Content and Presentation​

Scientific content is independent from visual layout.

The same manuscript may be rendered differently depending on publisher requirements without modifying its content.

Persistent Identity​

Every meaningful object possesses a persistent internal identifier.

Objects remain identifiable even if they move within the manuscript.

Traceability​

Every modification can be traced.

The history of each object forms part of the scholarly record.

Interoperability​

The model is designed to map cleanly onto open scholarly standards such as JATS XML and related metadata schemas.

The Manuscript Object

A manuscript is the root object of the document.

It contains:

  • metadata
  • contributors
  • document structure
  • scholarly objects
  • review information
  • publication history
  • relationships
  • version history
Manuscript
│
├── Metadata
├── Contributors
├── Sections
├── References
├── Figures
├── Tables
├── Notes
├── Files
├── Reviews
├── Versions
└── Publication

Semantic Objects

The manuscript consists of semantic objects.

Examples include:

  • Title
  • Subtitle
  • Abstract
  • Keyword
  • Author
  • Affiliation
  • ORCID
  • Section
  • Paragraph
  • Figure
  • Caption
  • Table
  • Equation
  • Citation
  • Bibliographic Entry
  • Footnote
  • Endnote
  • Comment
  • Review
  • Editorial Decision
  • Supplementary File

Each object possesses:

  • identifier
  • type
  • content
  • metadata
  • relationships
  • revision history

Relationships

Objects may reference other objects.

Examples:

  • citation → bibliography
  • author → affiliation
  • figure → caption
  • paragraph → note
  • review → manuscript version

The model therefore represents a graph rather than a simple linear document.

Text as an Object

Text itself is not the primary entity.

Instead, text is composed of semantic blocks.

Section

 ├── Heading
 ├── Paragraph
 ├── Paragraph
 ├── Figure
 ├── Paragraph
 ├── Table
 └── Paragraph

Anchors

Any semantic object may act as an anchor.

Annotations are attached to anchors rather than positions within rendered text.

Examples include:

  • comments
  • notes
  • citations
  • editorial remarks
  • AI suggestions

This allows annotations to remain valid even after substantial editing.

Flexible Notes

OMI generalizes the traditional concept of footnotes.

A note may be attached to:

  • a word
  • a phrase
  • a sentence
  • multiple sentences
  • an entire paragraph
  • a figure
  • a table
  • any semantic object

Rendering as footnotes, endnotes, sidenotes, inline notes, or other publisher-specific styles is determined during export.

References

Citations are independent semantic objects.

A citation references one bibliographic record.

Multiple citations may point to the same record.

Bibliographies are generated automatically.

Figures and Tables

Figures and tables are first-class scholarly objects.

They contain:

  • identifier
  • caption
  • source
  • licensing information
  • alternative text
  • linked files

Review Layer

Reviews are not embedded into the manuscript.

Instead, they form a separate annotation layer.

Different review workflows may expose different subsets of annotations.

Version Model

The manuscript evolves through versions.

Draft
    ↓
Submission
    ↓
Review
    ↓
Revision
    ↓
Accepted
    ↓
Published
    ↓
Corrected

Every version preserves complete provenance.

Publishing

Publication formats are generated from the semantic document.

Examples include:

  • JATS XML
  • HTML
  • PDF
  • EPUB
  • DOCX
  • Markdown

No publication format is considered canonical.

The semantic manuscript is the canonical source.

AI Interaction

AI systems never modify the manuscript directly.

Instead they create:

  • suggestions
  • annotations
  • metadata proposals
  • validation reports

Human approval is required before changes become part of the manuscript.

Future Extensions

The document model is intentionally extensible.

New semantic object types may be introduced without changing the overall architecture.

Examples include:

  • research data
  • executable notebooks
  • interactive figures
  • audiovisual material
  • software citations
  • research protocols

Guiding Principle

The manuscript is not a formatted document.

It is a structured scholarly knowledge object from which every presentation, workflow, and publication format can be derived.