chilon_rs

A fast, parallel Rust tool for namespace-based summarization of massive RDF graphs — extracts structure, infers namespaces, normalizes triples, and generates interactive visualizations.

Table of Contents

• Background
• Install
• Quick Start
• Usage
  • Basic Processing
  • CLI Options
  • Multiple Files
• Outputs & Visualization
• Auxiliary Commands
• Validation
• Citation
• Contributing
• License

Background

RDF graphs at web scale (hundreds of millions to billions of triples) are difficult to explore and understand. chilon_rs addresses this by:

1. Namespace Inference — automatically discovers namespace prefixes from IRIs using a community prefix table and statistical segmentation.
2. Triple Normalization — groups triples by inferred namespace, counts occurrences, and filters by minimum frequency to produce a compact summary.
3. Visualization — emits JSON data and a self-contained HTML/JS visualization for interactive exploration of the summary.

The algorithm is based on: dos Santos & Leal (2023). "Summarization of Massive RDF Graphs Using Identifier Classification." ICCS 2023.

Install

From Source (Recommended)

Requires Rust 1.77+.

git clone https://github.com/andrefs/chilon_rs.git
cd chilon_rs
cargo build --release

The binary will be at cli/target/release/chilon_rs.

Cargo Install (when published)

cargo install chilon_rs

Quick Start

# Build
cargo build --release

# Process an RDF file (Turtle format)
./cli/target/release/chilon_rs mygraph.ttl

This creates a dated folder under results/YYYYMMDD/ containing:

• Normalized triples grouped by namespace
• Namespace prefix table
• summary.json + visualization.html for interactive browsing

Usage

Basic Processing

# Process a single file with defaults (namespace inference ON, strict mode)
chilon_rs graph.ttl

CLI Options

Flag	Description
--no-infer-ns	Disable namespace inference; use only the built-in community prefix table.
-i, --ignore-unknown	Skip triples whose namespace cannot be resolved (instead of erroring).
-h, --help	Show help message.
-V, --version	Print version.

Multiple Files

# Process multiple files in one run
chilon_rs file1.ttl file2.ttl file3.ttl

Worker threads are auto-tuned: max(2, min(files + 1, CPU cores - 2)).

Outputs & Visualization

After processing, a folder results/YYYYMMDD-N/ is created with:

File	Description
output.ttl	Normalized triples grouped by namespace (Turtle).
vis-data.json	Visualization data (nodes, edges, aliases).
namespaces.tsv	Inferred + community prefix table (prefix, URI, count).
normalized.tsv	Normalized triples: subject\tpredicate\tobject\tcount.
tasks.json	Processing metadata (timings, triple counts, stages).
chilon.log	Detailed execution log.

Viewing the Visualization

The visualization is a JavaScript module that must be served over HTTP (browsers block ES modules on file://). Two steps:

1. Build the visualization assets (requires Node.js ≥ 18 + yarn):

   cargo run --bin gen-viz -- results/20260909-9

   This runs vite build inside chilon-viz/ and copies dist/ into the results folder.

2. Serve the dist/ folder and open in browser:

   cd results/20260909-9/dist && python3 -m http.server 8000
   # Then open http://localhost:8000

Auxiliary Commands

Two additional binaries ship with the crate:

# Generate visualization from an existing results folder
cargo run --bin gen-viz -- results/20260909-9

# Quick test/debug: parse RDF and show basic stats without full pipeline
cargo run --bin test-files -- file1.ttl file2.ttl

Validation

chilon_rs has been validated on 11 real-world RDF graphs spanning from a few MB / <1M triples to 90+ GB / billions of triples:

Dataset	Domain
ClaimsKG	Fact-checking claims
CrunchBase	Company/startup data
DbKwik	Wikipedia infobox extraction
DBLP	Bibliography / publications
DBpedia	Wikipedia structured data
KBpedia	Knowledge base ontology
LinkedMDB	Movie database
OpenCyc	Common-sense knowledge base
Wikidata	General knowledge graph
WordNet	Lexical database
YAGO	Ontology from Wikipedia

Summaries and visualizations: https://andrefs.github.io/chilon_rs

Citation

If you use chilon_rs or the underlying algorithm in your work, please cite:

dos Santos, A.F., Leal, J.P. (2023). Summarization of Massive RDF Graphs Using Identifier Classification. In: Ojeda-Aciego, M., Sauerwald, K., Jäschke, R. (eds) Graph-Based Representation and Reasoning. ICCS 2023. Lecture Notes in Computer Science. Springer, Cham. https://doi.org/10.1007/978-3-031-40960-8_8

Datasets

The 11 corpora used for validation are publicly available:

• ClaimsKG
• CrunchBase
• DbKwik
• DBLP
• DBpedia
• KBpedia
• LinkedMDB
• OpenCyc
• Wikidata
• WordNet
• YAGO

Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines on:

• Setting up the development environment (cargo test, cargo clippy --all-targets -- -D warnings)
• Code style and commit conventions
• Opening issues and pull requests

License

Released under the MIT License — see LICENSE for details.

Copyright (c) 2023 Alexandre F. dos Santos