SEMANTIC MODELS THAT RUN
========================

A fully functional, public demonstration is available at:

https://models-that-run.jargon.sh

-------------------------------------------------------------------------------

This submission demonstrates a model-driven, standards-aligned semantic
transformation pipeline. It generates JSON Schema, JSON-LD, and RDF-compatible
formats — all derived from a single source-of-truth domain model, created using
our company's Domain-Driven Design modelling platform: Jargon.

Jargon is used to model the logical structure of the information contained in
the OMG-provided CSV file, and also model any dependent domains (such as OGC
GeoSPARQL — which was imported into Jargon via their published ontology). From
this CSV domain model, Jargon generates the following artefacts, which have
been included in the submission:

- mapping.json : Declarative column-to-model mapping
- context.jsonld : Generated JSON-LD @context
- schema.json : Generated JSON Schema
- semantic-schema.json : JSON Schema that checks semantic alignment of the
  expanded JSON-LD

Jargon then plays no further role in the submission.

NOTE ON DATA SIZE:
   To ensure instant responsiveness in the live demo, only the first 100 rows
   of the full FDIC dataset are used. This subset is fully representative of
   the structure and semantics of the full dataset. The mappings,
   transformation engine, and generated artefacts are all capable of handling
   the complete dataset at scale, and have been tested offline for this
   purpose. The full dataset can be processed using the same declarative
   mapping and model-generated artefacts.
   The JSON output and expanded JSON-LD for the complete, 80,000 row sample are
   included in this ZIP, and described at the end of this document.



This submission is frozen as of May 28, 2025, in accordance with the competition
rules. No changes will be made to the live demo or underlying Jargon domain
model during the judging period. We reserve the right to make changes to both
the site and the Jargon domain model after the OMG competition winner is
announced and the competition has ended.

A standalone, read-only copy of the submission website (as it appeared at the
time of submission) is included in this ZIP as a PDF export. 
It provides evidence that the live demo has not changed post-deadline.

The entire transformation process runs live in the browser. It would run equally
well as part of a CI/CD pipeline or integrated into other developer tooling.

No backend logic, opaque scripting, or hardcoded output is used. Aside from
the artefacts generated by Jargon, the entire submission is driven by the CSV
data — which can be modified live in the demo.

Judges are encouraged to modify the input CSV directly and observe:

- Mapped JSON output using the declarative mapping
- JSON-LD expansion using the generated @context
- Structural validation using the JSON Schema
- Semantic alignment validation using the semantic-schema (using JSON Schema)
- RDF vocabulary output in Turtle and N-Quads formats

Each transformation step is dynamically derived from model-generated artefacts —
using standard libraries (with the exception of our CSV-to-JSON converter).

This ensures the solution is:

- Standards-compliant
- Transparent and reproducible
- Traceable from raw input to enriched output, with the model as a single source
  of truth

While the CSV-to-JSON engine used in the demo has not been submitted as source
code, a bundled and minified version is included in the submission website.

This tool (the converter, not Jargon) was developed specifically for this
challenge, and we have not yet determined whether it will be open-sourced or
commercialised.

However, the declarative mapping format used by the engine has been fully
described in mapping-readme.md, so its function can be understood, reused, or
re-implemented independently. The mapping approach has also been tested on
larger datasets and scales effectively by deduplicating references, nesting
reusable structures, and applying consistent rules derived from the model.

It doesn’t just add meaning to CSV data — it creates structured, semantically
rich outputs that are both intuitive for humans and directly usable by
developers. That challenge stubbornly persists with most semantic technology.

-------------------------------------------------------------------------------

Title for Submission: Semantic Models That Run
Primary Contact Name: Alastair Parker  
Email: alastair@jargon.sh  
Phone: +61 402 655 199 (an Australian phone number)
Organisation: Data Exchange Apparatus Company 
Website: https://jargon.sh

Copyright Waiver:  
All submitted materials remain the IP of the original owner.  
OMG may share any and all materials publicly.  
We intend to maintain and keep the demo running at its published URL:

https://models-that-run.jargon.sh

Included Files:

- SubmissionManifest.txt : This file  
- Semantic Models That Run.pdf : Static snapshot of the live website  
- mapping.json : Declarative column-to-model mapping  
- mapping-readme.md : Explanation of the mapping format  
- context.jsonld : Generated JSON-LD @context  
- schema.json : Generated JSON Schema  
- semantic-schema.json : Generated JSON Schema that checks semantic mapping  
- *example.json : Converted JSON from 80,000 line CSV  
- *example-expanded.jsonld : JSON-LD expanded form of example.json 
- *input.csv : OMG-provided FDIC dataset


* - these files are quite large
