WORK
WordNet
The computational lexical database Miller led the development of at Princeton from 1985 onward — a large-scale semantic network that organized English vocabulary into synonym sets linked by relations, and that became foundational infrastructure for natural language processing and, eventually, for the training of
large language models.
WordNet is a large-scale computational lexical database developed at Princeton University under
George Miller's leadership beginning in 1985. Its core innovation was organizing English vocabulary not as a traditional alphabetical dictionary but as a network of synonym sets (synsets) linked by semantic relations — hyponymy, meronymy, antonymy, and others. Each synset represents a distinct concept; the links
between synsets represent the conceptual structure of the lexicon. WordNet grew into a resource of over 117,000 synsets covering nouns, verbs, adjectives, and adverbs, and became foundational infrastructure for natural language processing research across decades. Its design reflected Miller's lifelong interest in how humans organize meaning — the same interest that had drawn him from memory research to language research in the 1960s. WordNet was, in a sense, the computational instantiation of the chunking hierarchy Miller had been theorizing for three decades: a systematic mapping of the compressed categories through which human