dl-ontologies-influence-conceptual-similarity

Summary

The way a system formally encodes concept relationships (like a structured taxonomy or ontology) directly shapes how it judges whether two ideas are "similar," so similarity scores are not objective readings of language but reflections of the underlying knowledge model you chose. This means any NLP pipeline relying on similarity metrics is partly an expression of its ontology, and swapping or extending that ontology will shift downstream results like retrieval, classification, or matching.