An Abstract Multilingual WordNet

Download PDF

Krasimir Angelov

Published: 29 Jan 2025, Last Modified: 20 Feb 2025OpenReview Archive Direct UploadEveryoneCC BY 4.0
Abstract: We present a variant of WordNet for 265 languages where the primary constituents of the synsets are abstract identifiers, rather than language specific lexemes. The identifiers are then verbalized to each language through a grammar. Currently, for most of the languages, the grammar only provides lemmas, but for 28 of them, there is also, full morphology and syntax. We review the bootstrapping methodology, evaluate the quality, and show-case applications.