NBA Knowledge Graph Project

Course: Bases de Données Spécialisées – Master 2 MIADS
University: Université Paris Cité
Academic Year: 2025-2026
Authors: Chris ESSOMBA, Ivan BANFOU

Overview

This project implements a knowledge graph focused on the NBA (players, teams, and games) as part of the "Knowledge Graphs et Raisonnement" assignment.

The goal is to demonstrate the full pipeline:

• Extraction and integration of heterogeneous open datasets
• Conversion to RDF
• Advanced SPARQL querying
• RDFS reasoning with inference

Datasets

Two public datasets from Kaggle were used:

1. NBA Games – Nathan Lugan
   Link: https://www.kaggle.com/datasets/nathanlauga/nba-games
   Contains match details (date, teams, scores, season, etc.).

2. NBA Players & Teams – Wyatt Walsh
   Link: https://www.kaggle.com/datasets/wyattowalsh/basketball
   Contains player information and current team assignments.

Common concept: team_id → used for integration.

Project Structure

.

├── data/
│   ├── Games.csv
│   └── Players.csv
├── turtles_files/
│   ├── nba_graph.ttl                # Integrated RDF data
│   └── schema_rdfs_complet.ttl      # RDFS ontology
├── notebook.ipynb                   # Complete workflow (extraction, RDF generation, queries, reasoning)
└── README.md                        # This file

Key Features

• RDF Modeling: Unique URIs for players, teams, and games. Links via playsFor, homeTeam, visitorTeam.
• RDFS Schema: Hierarchical classes (Player ⊑ Person, Team ⊑ Organization, Game ⊑ Event) and properties (involvesTeam, hasName, hasPoints) with subPropertyOf, domain/range.
• 10 SPARQL Queries: Including federation (DBpedia), OPTIONAL, named graphs, aggregation, property paths, MINUS/FILTER NOT EXISTS.
• RDFS Reasoning: Using owlrl to show inferences (subClassOf, subPropertyOf, domain/range).

How to Run

1. Install dependencies:

pip install -r requirements.txt