Expertise Knowledge Graph

These are my notes that together form a semantic knowledge graph built with Logseq. They document my professional expertise, projects, skills, and career history. The knowledge graph is automatically exported to RDF, validated using SHACL shapes, and published as a static website, PDF CV and RDF data. The RDF data is available both as ttl serialization and through SPARQL endpoint.

🌐 Live Deployments

• Website: expertise.matdata.eu
• SPARQL Endpoint: jena.mkwadraat.be/expertise
• CV (PDF): expertise.matdata.eu/static/Mathias_Vanden_Auweele_CV.pdf
• RDF Data:
  • matdata-expertise.ttl (cleaned & validated)
  • matdata-expertise-raw.ttl (raw export)
  • matdata-expertise-shacl.ttl (validation shapes)

📖 Overview

This repository contains my personal knowledge management system, structured as a semantic knowledge graph using Logseq as the authoring environment. The graph models:

• Projects: Professional work and initiatives
• Techniques & Tools: Technologies, programming languages, frameworks
• Roles: Professional roles and responsibilities
• Jobs: Employment history with timeline
• Companies: Organizations I've worked with
• Talks & Presentations: Public speaking engagements
• Categories: SKOS-based taxonomies for organizing expertise

All content is authored in Logseq markdown pages with structured properties, automatically converted to RDF using the logseq-rdf-export tool, and validated against SHACL shapes to ensure data quality.

🏗️ Architecture

Workflow

Logseq Pages (Markdown)
    ↓
logseq-rdf-export
    ↓
matdata-expertise-raw.ttl
    ↓
clean-graph.py (normalization)
    ↓
matdata-expertise.ttl
    ↓
validate-graph.py (SHACL validation)
    ↓
├─→ Static Website (Logseq SPA)
├─→ SPARQL Endpoint (Apache Jena Fuseki)
└─→ CV Generation (RenderCV)

Key Components

1. Logseq: Knowledge authoring with structured properties
2. RDF Export: Custom tool to convert Logseq graph to RDF/Turtle
3. Graph Cleaning: Python script to normalize predicates and map categories
4. SHACL Validation: Ensures data quality and consistency
5. CI/CD Pipeline: GitLab/GitHub CI automates the entire workflow
6. Publishing: Multi-channel output (website, SPARQL, PDF CV)

🚀 Local Development

Prerequisites

• Python 3.12+
• Docker (for RDF export)
• Git

Setup

1. Clone the repository:

git clone <repository-url>
cd Expertise

2. Create a Python virtual environment:

python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\Activate.ps1
pip install -r kg-actions/requirements.txt

Local Validation Workflow

Run the complete validation workflow locally:

# 1. Export Logseq to RDF
docker run -it -v ./:/data mathiasvda/logseq-rdf-export logseq-rdf-export matdata-expertise-raw.ttl --directory /data/

# 2. Clean and normalize the graph
python ./kg-actions/clean-graph.py

# 3. Validate against SHACL shapes
python ./kg-actions/validate-graph.py

Expected output for successful validation:

Number of triples in data graph: 2900
Number of triples in shacl graph: 168
Validation Report
Conforms: True

Working with Logseq

Edit pages in the pages/ directory following these conventions:

Projects:

public:: true
type:: [[Project]]
description:: Brief description
has-category:: Strategy
has-tagged-techniques:: #Python, #RDF
has-tagged-roles:: #Developer
is-featured:: Yes
during-job:: #[[Job: Independent railway data freelancer]]

Techniques:

public:: true
type:: [[Technique]]
self-estimated-proficiency:: Proficient
is-featured:: Yes
has-category:: Programming languages

See kg-actions/shapes-explanation.md for complete validation rules.

🔄 CI/CD Pipeline

The GitLab CI pipeline consists of 5 stages:

1. RDF Export: Convert Logseq to RDF using Docker image
2. Validate: Clean graph and run SHACL validation
3. CV: Generate PDF CV using RenderCV
4. Pages: Build static website using Logseq Publish SPA
5. Upload: Sync RDF data to SPARQL endpoint

Pipeline runs automatically on commits to the main branch.

📊 Data Model

The knowledge graph uses these main entity types:

• d:Project - Professional projects
• d:Technique - Skills, tools, technologies
• d:Role - Professional roles
• d:Job - Employment history
• d:Company - Organizations
• d:Talk - Presentations and talks
• skos:Concept - Taxonomy categories
• skos:ConceptScheme - Category schemes

Key predicates:

• d:has-category - Link to SKOS category
• d:has-tagged-techniques - Technologies used in project
• d:has-tagged-roles - Roles performed in project
• d:self-estimated-proficiency - Skill level
• d:is-featured - Highlight important items
• d:during-job - Job context for projects

🎯 SHACL Validation

SHACL shapes enforce:

• Required properties (labels, categories, dates)
• Cardinality constraints (min/max counts)
• Data type validation (strings, booleans, dates, IRIs)
• Category membership in correct SKOS concept schemes
• Proficiency level consistency
• Relationship integrity

See kg-actions/matdata-expertise-shacl.ttl for complete shapes.

📝 Scripts

• clean-graph.py: Normalizes RDF graph by mapping properties and merging SKOS concepts
• validate-graph.py: Validates RDF graph against SHACL shapes
• formatter.py: Utilities for RDF formatting
• complete-cv.py: Generates CV YAML from RDF data for RenderCV

🔗 Related Projects

• logseq-rdf-export - Export tool for Logseq to RDF
• Yasgui - SPARQL GUI for querying the endpoint
• RenderCV - CV generation from YAML