WordNet is a large lexical database of English begun by the psychologist George A. Miller at Princeton University and described in his 1995 Communications of the ACM article “WordNet: A Lexical Database for English.” It organizes nouns, verbs, adjectives, and adverbs into “synsets” - sets of synonyms that each stand for one distinct concept - and then links those concepts through semantic relations: hypernymy and hyponymy (is-a hierarchies), meronymy and holonymy (part-whole), antonymy, and others. The result is part dictionary, part thesaurus, and part hand-built knowledge graph of meaning.
For decades WordNet was foundational infrastructure for natural language processing. Word-sense disambiguation, information retrieval, and many early semantic systems leaned on it, and it offered a structured, human-curated map of vocabulary at a time when statistical methods had little notion of meaning. It is a flagship example of the symbolic, knowledge-engineering tradition in AI - intelligence built by careful human encoding rather than learned from data.
WordNet’s most consequential downstream role came in computer vision. When Fei-Fei Li and colleagues built ImageNet starting in 2007, they used WordNet’s noun hierarchy as the scaffold for organizing categories - populating WordNet concepts with labeled images. WordNet thus connects the oldest hand-built-knowledge thread of AI to the dataset that helped ignite the deep-learning era. For business readers, it shows how a patiently assembled reference resource can quietly enable breakthroughs decades later in fields its authors never targeted.