🧭 A system map for the Federal Office for Agriculture

This repository is set up to gather and visualize information about IT systems, the data those contain and their operating organizations in the Swiss agri-food sector. You can visually explore this work on the following pages:

• DigiAgriFood system map, a visualization of the system map knowledge graph that specifically shows the classes organization, system and information unit.
• Circular system map, an alternative visualization in circular arrangement.
• Ontology visualization made with WebVOWL, allowing to get a quick grasp of the ontology underlying the system map.

If instead you are interested in the actual data, you can have a look at the following files:

• ontology.ttl contains information about how the data is structured.
• data.ttl contains information about the organizations, systems and information units themselves.

Actually, there is a python script reasoning over the two files. This is described in more detail below.

⚙️ Setting query parameters in the system map visualization

In the DigiAgriFood system map, we can set URL parameters to change the language or the underlying LINDAS SPARQL query. For example, you might want to visualize only IT systems and the data flow for the Swiss agri-food sector, without any organizational information, but in French. You can do that with:

https://blw-ofag-ufag.github.io/system-map/index.html?lang=fr&organization=false

Here's a table of the possible URL parameters, descriptions and their default values.

Parameter	Possible values	Default	Description
lang	de, fr, it, en	de	Defines the language for labels and descriptions. Currently, the graph contains some langstrings in German, French, Italian and English. If the node or edge label is not found in said language, a question mark is shown instead.
organization	true, false	true	Determines whether organizations (schema:Organization) should be displayed.
system	true, false	true	Determines whether IT systems (schema:SoftwareApplication) should be displayed.
information	true, false	true	Determines whether information units (systemmap:Information) should be displayed.

🔎 Examples of SPARQL queries from LINDAS

• Is there (sensitive) personal data in a system?
• Table of Cantons, their agricultural IT-Systems and the respective operating agency (if given)

💭 Reasoning with python

The repository includes a Python script scripts/reason.py that processes the two Turtle files (data.ttl and ontology.ttl) by first sorting them and then applying custom reasoning before producing the final output (graph.ttl). Here's how it works:

1. Sorting the turtle files: Before any reasoning is applied, both ontology.ttl and data.ttl are loaded and sorted using a custom serializer (based on otsrdflib). This ensures that:

   • The triple order is consistent for easier human reading and debugging.
   • The versioning of the files in GitHub stays more consistent.
   • The syntax stays consistent (e.g., new line for each triple etc.)

2. Merging and custom reasoning: The script then merges the ontology and data graphs into a single RDF graph and performs iterative reasoning based on two primary rules:

   • Subclass reasoning: If an instance is declared as being of a certain class A, and the ontology specifies that A is a subclass of another class B (using rdfs:subClassOf), the script infers that the instance is also of type B.
   • Inverse Property Reasoning: If a property p is defined to have an inverse q (using owl:inverseOf), then for every triple (s, p, o), the script adds the inverse triple (o, q, s) to the graph.

   This reasoning is applied iteratively—repeating the process until no new triples are added—so that all valid inferences are captured in the final graph.

3. Final output: Once reasoning is complete, the enriched graph is saved to graph.ttl. The output file is sorted as well, ensuring consistency and readability. This final sorted and reasoned graph is the one that is pushed to LINDAS.

This processing pipeline enhances the raw data with logical inferences based on the ontology and hence improves the consistency of the produced graph.

🆙 Guide to upload the turtle files to LINDAS

In order to upload a graph (as a turtle .ttl file) to the linked data service LINDAS, use the following cURL command:

curl \
  --user lindas-foag:mySuperStrongPassword \
  -X POST \
  -H "Content-Type: text/turtle" \
  --data-binary @graph/plant-protection.ttl \
  "https://stardog-test.cluster.ldbar.ch/lindas?graph=https://lindas.admin.ch/foag/ontologies"

Replace mySuperStrongPassword with the actual password. Of course, graph/plant-protection.ttl could be set to another turtle file and https://lindas.admin.ch/foag/ontologies to another target graph.

If data that was already uploaded was changed, clear the existing graph before posting. (Otherwise, stardog creates duplicate nodes for the changes.) To clear the graph, run:

curl \
  --user lindas-foag:mySuperStrongPassword \
  -X DELETE \
  "https://stardog-test.cluster.ldbar.ch/lindas?graph=https://lindas.admin.ch/foag/ontologies"