Reproduce “OntoLearner: A Modular Python Library for Ontology Learning with Large Language Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Hamed Babaei Giglou, Jennifer D'Souza, Andrei Aioanei, Nandana Mihindukulasooriya, Sören Auer
OntoLearner is a modular Python library that integrates ontology access, large language model (LLM)-driven learning, and standardized benchmarking, establishing an infrastructure for reproducible research in neuro-symbolic ontology engineering. An empirical study with the library reveals that failures in ontology learning are predominantly determined by the ontology's structure rather than the model's architecture.
Key insights
- The primary bottlenecks in Ontology Learning (OL) are ontology-determined structural mismatches with model knowledge encoding, rather than solely model architecture limitations.
- Taxonomy discovery presents a dominant challenge, as existing models struggle with global hierarchical structure inference and specific entailment relationships, even with increased scale.
- For structured extraction tasks, LLMs benefit more from instruction tuning for schema compliance than from raw reasoning depth, while current embeddings exhibit anisotropy, limiting their ability to differentiate nuanced taxonomic relations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.