Ontolearn-A Framework for Large-scale OWL Class Expression Learning in Python

TL;DR

Ontolearn addresses the need for scalable, explainable OWL class expression learning on web-scale RDF knowledge graphs. It unifies nine modern learners (symbolic, neuro-symbolic, and deep) including EvoLearner and DRILL, provides DL-Learner wrappers, SPARQL-backed querying to remote triplestores, and an LLM-based verbalization module. Learning problems are defined over a knowledge base with positive $E^+$ and negative $E^-$ examples, enabling the extraction of DL expressions that classify instances. The framework is open-source (MIT), well-tested (156 tests, 95% coverage), widely adopted (over 26k downloads), and supports industrial deployment via robust triplestore integration and reasoning.

Abstract

In this paper, we present Ontolearn-a framework for learning OWL class expressions over large knowledge graphs. Ontolearn contains efficient implementations of recent stateof-the-art symbolic and neuro-symbolic class expression learners including EvoLearner and DRILL. A learned OWL class expression can be used to classify instances in the knowledge graph. Furthermore, Ontolearn integrates a verbalization module based on an LLM to translate complex OWL class expressions into natural language sentences. By mapping OWL class expressions into respective SPARQL queries, Ontolearn can be easily used to operate over a remote triplestore. The source code of Ontolearn is available at https://github.com/dice-group/Ontolearn.

Figures (2)

• Figure 1: Ontolearn architecture. Green rectangles show the top-level components, whereas beige rectangles show the sub-components.
• Figure 2: A partial visualization of the Family knowledge base along with a learning problem defined by $E^+$ and $E^-$. An example of a learned concept is given in DL syntax which is verbalized into natural language using an LLM.