Loreto

Loreto stands for Logic Ontology Representation Tokenization Optimisation.

Loreto is a compact textual serialization for RDF/OWL datasets designed to reduce token consumption for large language models while preserving the underlying graph structure needed for ontology use, retrieval tasks, and reasoning-oriented workflows.

Why Loreto

Standard RDF serializations such as RDF/XML, Turtle, N-Triples, TriG, and JSON-LD are useful for interoperability, but they are not optimized for LLM prompt efficiency.

Loreto focuses on:

• compact namespace planning
• compact repeated IRI representation
• low syntactic overhead
• LLM-friendly graph layout
• preservation of literal lexical forms
• preservation of named graphs and dataset contexts

Quick Start

Install the core converter dependencies:

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Convert an ontology to Loreto:

python3 loreto.py input.owl
python3 loreto.py input.ttl --no-ttl
python3 loreto.py input.rdf -o output.loret

The converter writes:

• input.loret
• optionally input.normalized.ttl

Repository Contents

• loreto.py: main RDF/OWL to Loreto converter
• benchmark/: benchmark suite covering size, model performance, and reasoning
• examples/: small RDF, OWL, and SWRL examples with Loreto outputs
• docs/: technical notes on the format and benchmarks
• requirements.txt: core runtime dependencies

Repository Structure

Loreto/
├── README.md
├── LICENSE
├── loreto.py
└── examples/
    ├── README.md
    ├── basic/
    ├── owl/
    └── swrl/

Examples

Small concrete examples are available in examples/:

• examples/basic/ for a minimal RDF graph
• examples/owl/ for OWL content encoded as RDF
• examples/swrl/ for SWRL represented through RDF triples