Paris 2024 Olympics - Semantic Web Project

A comprehensive knowledge graph project for the Paris 2024 Olympic Games, combining semantic web technologies with advanced NLP for data enrichment.

📋 Overview

This project builds an enriched semantic model of Olympic Games data through:

• Custom RDF/OWL ontology for Olympic domain modeling
• Deep learning-based information extraction using spaCy's fr_core_news_lg
• SKOS-based sport classification system
• SPARQL inference rules and SHACL constraints
• External knowledge integration (DBpedia/Wikidata)
• Real-time weather data integration

🧠 Deep Learning & NLP

The project leverages spaCy's fr_core_news_lg model (1.7GB) for advanced text analysis:

• Named Entity Recognition optimized for athlete identification
• Sport-specific term classification
• Performance metric extraction
• Contextual relationship mapping

Model Performance:

• 89% accuracy on athlete name recognition
• 97% accuracy on sports terminology classification
• 2,500+ athlete mentions processed
• 1,800+ performance records extracted

🏗️ Architecture

Semantic Layer

• Modular ontology with Person, SportingEvent, and Location hierarchies
• SKOS taxonomy for sports classification
• SHACL constraints for data validation
• Custom SPARQL rules for knowledge inference

Data Integration Layer

• DBpedia and Wikidata entity linking
• Real-time weather data integration
• Unstructured text processing pipeline
• CSV data transformation system

Visualization Layer

• Interactive knowledge graph exploration
• Medal distribution analytics
• Event timeline visualization
• Weather condition monitoring

📈 Visualizations

Knowledge Graph (KG)

Medals queries

Simple charts

🛠️ Prerequisites

• Python 3.8+
• Docker 20.10+
• 4GB RAM minimum (8GB recommended for full graph processing)

🚀 Quick Start

1. Clone and install dependencies:

git clone https://github.com/[username]/jo2024-semantic.git
cd jo2024-semantic
pip install -r requirements.txt

2. Launch weather service:

docker compose up -d

3. Run data pipeline:

python 01_data_enricher.py  # NLP-based enrichment
python 02_data_linking.py   # Knowledge base linking
python 03_generate_graph.py # Visualization generation

📊 Key Results

• Initial dataset: 355 triples
• After enrichment: 15,152 triples
• External links created: 2,843
• NLP-extracted relationships: 4,500+
• Real-time weather monitoring for 5 venues

🔍 Available Analysis

1. Athlete Performance Analysis

   • Medal distribution by country
   • Performance trends over time
   • Cross-discipline achievements

2. Event Analysis

   • Temporal distribution
   • Venue utilization
   • Weather impact assessment

3. Knowledge Graph Exploration

   • Entity relationship visualization
   • Path finding between entities
   • Cluster analysis

⚠️ Known Limitations

• Weather API coordinate precision limited by infrastructure constraints
• Graph visualization limited to 100 triples for performance
• French language model occasionally struggles with rare athlete names

🗺️ Future Development

• Expansion of the sports ontology
• Integration of live event streaming data
• Enhanced predictive analytics
• Multi-language support

✍️ Authors

• Marc Pinet - marcpinet
• Arthur Rodriguez - rodriguezarthur
• Amine Haddou - H4znow

📄 License

MIT License - see LICENSE file for details.

For detailed documentation on SPARQL queries and ontology structure, see the docs/ directory.