# Ontology Learning **Domain:** Artificial Intelligence / Knowledge Representation / Natural Language Processing **Doc Type:** Canonical Concept Node **Maturity:** Developed ## Definition **Ontology learning** is the automatic, semi-automatic or human-guided discovery of concepts, classes, relations and sometimes axioms from text, structured data or repeated domain usage. It converts recurring patterns in a corpus into candidate knowledge structures. Ontology learning is also called ontology induction in some contexts. It may use terminology extraction, clustering, taxonomic analysis, relation extraction, statistical association, language models and expert review. ## Corpus Context Ontology learning describes the bottom-up phase of [[wiki/Bidirectional Ontology|Bidirectional Ontology]]. Articles repeatedly encounter a mechanism; the recurrence reveals a conceptual object; comparison exposes its boundaries; and the object becomes a wiki node linked back into the corpus. The process is evidentiary rather than merely lexical. Repeated words do not automatically justify a concept, and different words may describe the same underlying mechanism. Candidate nodes must be tested against existing definitions, neighboring concepts and the purposes for which the distinction will be used. ## Human and Machine Roles Computational systems can identify repeated terms, co-occurrences, clusters and probable relations at scales difficult for a human reader. Human judgment remains necessary to determine whether the pattern is substantive, whether its name is appropriate and whether its inclusion would clarify or distort the ontology. Language models can assist with candidate discovery and comparison, but fluent synthesis must not be mistaken for validated conceptual structure. Provenance and evidence boundaries remain part of the learned ontology. ## Key Insight **A corpus can teach an ontology what distinctions it needs, but frequency alone cannot decide which distinctions deserve authority.** ## Sources / Provenance - Applied to this corpus in [[journal/2026-08-15 - Bidirectional Ontology and Semantic Elasticity|2026-08-15 — Bidirectional Ontology and Semantic Elasticity]]. - A recognized area of ontology engineering and natural-language processing, commonly associated with term, concept, taxonomy, relation and axiom extraction. ## See Also [[wiki/Ontology|Ontology]], [[wiki/Bidirectional Ontology|Bidirectional Ontology]], [[wiki/Ontology Evolution|Ontology Evolution]], [[wiki/Knowledge Representation|Knowledge Representation]], [[wiki/Corpus Engineering|Corpus Engineering]], [[wiki/Semantic Network|Semantic Network]], [[wiki/Provenance|Provenance]]