Dictionary-guided sense projection expands WordNet-style resources to new languages
University of Alberta researchers describe a project-and-filter method that attaches target-language words to existing lexical concepts, releasing code and generated sense inventories.
In plain words
WordNet-style resources link words to shared concepts, but many languages lack them. The team's method takes sense-tagged English text and its translation, then projects the concept labels onto matching words in the other language. A bilingual dictionary steers the alignment and filters bad projections, improving precision without heavy computing.
What was announced
The paper, an arXiv abstract from the UAlberta-NLP group, frames the task as automatically expanding WordNet-style lexical resources to new languages through sense generation. Given a sense-tagged English corpus and its translation, the method projects annotated synsets onto aligned target-language tokens and assigns the corresponding lemmas to those synsets. To build and quality-check the alignments, the authors augment a pretrained base aligner with a bilingual dictionary, which also filters incorrect sense projections.
The abstract reports evaluation across multiple languages against prior methods plus dictionary-based and large language model baselines. The proposed project-and-filter strategy is said to improve precision while remaining interpretable and resource-efficient. The authors release code, documentation, and the generated sense inventories on GitHub. The abstract does not name the languages tested, give figures, or report results beyond the authors' own comparison, so independent confirmation is not available from this source.
What it does not show
- The source is only the paper's abstract; it names no specific languages, metrics, or numerical results, and the claims are the authors' own without independent verification.
For security and IT teams
- Access
- Code, documentation, and generated sense inventories are publicly released at the UAlberta-NLP ExpandNet GitHub repository.
Sources
Source-support review completed 17 Nov 2026. This story was drafted with AI assistance and checked against retained source passages. Automated checking is not independent verification.
Coverage: arXiv: abstract.
Updates and corrections
No corrections so far.
First published 17 Nov 2026
New to the subject? AI from the Ground Up starts from the beginning.