Small Knowledge Graphs (small-kgs)

A collection of small knowledge graphs in multiple formats for graph ML research and development.

Dataset Structure

This dataset contains knowledge graphs in three formats:

graph-std

Compressed Sparse Row (CSR) graph format with parquet files containing:

• Node data (persons, organizations, events, concepts, places, knowledge graphs)
• Edge indices and indirection pointers
• Node/edge type mappings
• Schema definition

duckdb

DuckDB database format for efficient analytical queries on graph data.

lbdb

LadybugDB format for graph database operations using the LadybugDB embedded graph database.

Usage

from datasets import load_dataset

# Load graph-std version
dataset = load_dataset("ladybugdb/small-kgs", name="graph-std")

# Load duckdb version  
dataset = load_dataset("ladybugdb/small-kgs", name="duckdb")

# Load lbdb version
dataset = load_dataset("ladybugdb/small-kgs", name="lbdb")

Contents

Version	File	Size
graph-std	csr_graph_*.parquet	~200KB
duckdb	kg_history.duckdb	~5MB
lbdb	kg_history.lbdb	~8MB