Christiane Fellbaum (Princeton University): Mapping the lexicon with WordNet

The lexicon of a language is vast, irregular, and open-ended; yet human language users effortlessly store and retrieve tens of thousands of words along with knowledge about the concepts behind the words.

Psychological and textual evidence indicates that meaning similarity is a major principle in the organization of the lexicon. While
traditional dictionaries do not reflect this, the digital resource WordNet interconnects words based on their semantic relatedness.
The resultant network can be navigated by computers to detect and measure meaning similarity and WordNet's use in natural language understanding provides a testing ground for theories of the lexicon and lexicalization patterns.

WordNet assumes stable, discrete and enumerative word senses, a premise that is challenged by experiments with manual semantic annotation and automatic word sense disambiguation. We examine some of the evidence against static meaning representation and examine alternatives.