First public release in preparation
GalleyForge – project
GalleyForge turns accepted manuscripts into structured data and clean HTML articles for small scientific journals. This page describes the problem, the current state of development, my role, one worked example and a project from practice. Nothing here is a service offer.
Problem
In small journals, production still means manual work between the accepted Word file and the published article: structuring sections, placing figures and tables, formatting references, entering metadata and checking the result. The effort recurs for every article, differs between staff members and is rarely documented. Larger publishers solve this with proprietary systems that small journals cannot afford.
Development status
Every item below was checked against the repository code and its automated tests. “Prototype” means it runs, but is not production-ready.
Already in the prototype
| Capability | What exists | Limits |
|---|---|---|
| DOCX to structured data to HTML | A Python prototype reads a DOCX manuscript, detects title, authors, affiliations, corresponding author, keywords, sections, tables and figure legends, writes a versioned JSON schema and renders a basic HTML article. Source code: GalleyForge Core on GitHub (MIT licence, fictional example data only). | Heuristics that rely on layout conventions. Manuscripts with other layouts (for example comma-separated authors without labels, or introductory lines) are not handled reliably. Tables and figures are collected in their own blocks, not at their place in the text; references stay plain text. Small test set with fictional data only. No PDF or XML output. |
| Rule-based validation of the structured data | Business rules report errors, warnings and hints (for example missing keywords or author data). | A starting set of rules. |
| Platform backend (foundation) | Sign-in and sessions, organisations and journal-specific roles, journal settings, article records, file upload and storage, structured article content with conflict-safe saving, and an endpoint that validates article content (two rules so far). Covered by 503 automated tests. | Backend only: there is no browser interface yet, no import or export in the platform, no deployment. |
| Journal-specific internal tools | Scripts and a small desktop tool for generating HTML articles, and a separate generator for PubMed XML metadata, built for the author's own journal production. | Specific to one journal, not generalised, no automated tests. They inform the product design but are not part of it. |
Planned for the first version
- Browser-based review editor with autosave (the saving contract is designed and tested on the backend side)
- DOCX import connected to the platform
- HTML export of a finished article, including figures and tables
- Journal-level configuration of required content, enforced by validation
Later roadmap
- PDF generation
- Crossref XML and PubMed XML export from the platform
- Complete publication package (ZIP)
- Journal templates, validation profiles and export profiles for several journals
- AI-assisted structure analysis – always optional and under editorial control
No release date has been set.
My role
I initiated GalleyForge and am responsible for the product requirements and the editorial workflows. I bring in the requirements from publishing practice, define the editorial processes, prioritise functions and review the development. More about my background is on the About page.
Example workflow
The example uses a fictional manuscript created for this page. No real article, author or journal data is involved.
- Input: a DOCX file with a notice banner, title, three authors with superscript affiliation labels, two affiliations, a correspondence line, keywords, the usual sections, a declarations block, one table with caption and one figure legend.
- Structuring: the prototype writes a JSON document: title, authors with affiliation indices, affiliations, corresponding author, keywords, sections, table rows and figure legend.
- Output: a basic HTML article including the table and the figure legend, generated without any manual editing.
Open the sample output (unedited generator output)
Limits that remain
- All elements of this fictional test manuscript were extracted correctly (title, authors with affiliation links, correspondence, keywords, sections, table, figure legend). The test manuscript was written for this purpose; real manuscripts vary much more.
- Visible limits in the sample: the table and the figure legend appear in their own blocks after the main text, not where the text mentions them; the note “No external references” appears as a reference entry; no image is attached; the page is plain, without journal layout, metadata or PDF.
- Earlier versions of the prototype misread this kind of layout (correspondence line taken as an author, keywords missing, no tables or legends). The parser was extended and regression tests were added. Layouts with comma-separated authors, introductory lines or unusual headings are still not handled.
Manuscripts that break these conventions are exactly what the planned review step is for: the editor corrects the structured data before the export, and validation points to the gaps.
Work in progress
Project from practice: a web-based multi-journal submission and peer-review system
A separate project from my work as managing director of a publishing house; it is not part of GalleyForge.
Starting point
Submission and peer review for several journals are to be handled in one web-based system with journal-specific settings, replacing previous manual routines.
My role
I initiated the project; its requirements and editorial processes come from my publishing practice.
Areas of focus
- Manuscript versions: revisions of a manuscript are to stay traceable through review and author revision.
- Reviewer management: management of reviewer data, including contact history.
- Reviewer search: support for editors in finding suitable reviewers, with indications of possible conflicts of interest; the decision is to stay with the editors.
- Invitations and communication: invitations and further steps are to follow defined, traceable processes.
Development
The system is being developed further on an ongoing basis.
Feedback and pilot interest
Questions, feedback from editorial practice or interest in a later pilot: hello@galleyforge.com. Please do not send manuscripts or personal data.