SciKnowOrg/ontolearner-law

Law Domain Ontologies

Overview

The law domain encompasses ontologies that systematically represent the complex structures and interrelations of legal concepts, processes, regulations, and rights. This domain is pivotal in knowledge representation as it facilitates the formalization and interoperability of legal information, enabling precise reasoning and decision-making across diverse legal systems. By capturing the intricacies of legal language and practice, these ontologies support the automation and enhancement of legal services and research.

Ontologies

Ontology ID	Full Name	Classes	Properties	Last Updated
CopyrightOnto	Copyright Ontology (CopyrightOnto)	38	12	2019-09

Dataset Files

Each ontology directory contains the following files:

1. <ontology_id>.<format> - The original ontology file
2. term_typings.json - A Dataset of term-to-type mappings
3. taxonomies.json - Dataset of taxonomic relations
4. non_taxonomic_relations.json - Dataset of non-taxonomic relations
5. <ontology_id>.rst - Documentation describing the ontology

Usage

These datasets are intended for ontology learning research and applications. Here's how to use them with OntoLearner:

First of all, install the OntoLearner library via PiP:

pip install ontolearner

How to load an ontology or LLM4OL Paradigm tasks datasets?

from ontolearner import CopyrightOnto

ontology = CopyrightOnto()

# Load an ontology.
ontology.load()  

# Load (or extract) LLMs4OL Paradigm tasks datasets
data = ontology.extract()

How use the loaded dataset for LLM4OL Paradigm task settings?

# Import core modules from the OntoLearner library
from ontolearner import CopyrightOnto, LearnerPipeline, train_test_split

# Load the CopyrightOnto ontology, which contains concepts related to wines, their properties, and categories
ontology = CopyrightOnto()
ontology.load()  # Load entities, types, and structured term annotations from the ontology
data = ontology.extract()

# Split into train and test sets
train_data, test_data = train_test_split(data, test_size=0.2, random_state=42)

# Initialize a multi-component learning pipeline (retriever + LLM)
# This configuration enables a Retrieval-Augmented Generation (RAG) setup
pipeline = LearnerPipeline(
    retriever_id='sentence-transformers/all-MiniLM-L6-v2',      # Dense retriever model for nearest neighbor search
    llm_id='Qwen/Qwen2.5-0.5B-Instruct',                        # Lightweight instruction-tuned LLM for reasoning
    hf_token='...',                                             # Hugging Face token for accessing gated models
    batch_size=32,                                              # Batch size for training/prediction if supported
    top_k=5                                                     # Number of top retrievals to include in RAG prompting
)

# Run the pipeline: training, prediction, and evaluation in one call
outputs = pipeline(
    train_data=train_data,
    test_data=test_data,
    evaluate=True,              # Compute metrics like precision, recall, and F1
    task='term-typing'          # Specifies the task
                                # Other options: "taxonomy-discovery" or "non-taxonomy-discovery"
)

# Print final evaluation metrics
print("Metrics:", outputs['metrics'])

# Print the total time taken for the full pipeline execution
print("Elapsed time:", outputs['elapsed_time'])

# Print all outputs (including predictions)
print(outputs)

For more detailed documentation, see the

Citation

If you find our work helpful, feel free to give us a cite.

@inproceedings{babaei2023llms4ol,
  title={LLMs4OL: Large language models for ontology learning},
  author={Babaei Giglou, Hamed and D’Souza, Jennifer and Auer, S{\"o}ren},
  booktitle={International Semantic Web Conference},
  pages={408--427},
  year={2023},
  organization={Springer}
}