FEWS (Few-shot Examples of Word Senses) is a few-shot dataset for English Word Sense Disambiguation (WSD) gathered from Wiktionary, an online, crowd-sourced dictionary. FEWS contains over 121,000 labeled examples of ambigous words, corresponding to more than 71,000 sense types. The evaluation for FEWS is split into few-shot and zero-shot settings, to better faciliate evaluating on few-shot learning and perfromance on rare senses.
We provide a data sheet summarizing the dataset and outlining intended uses of the data. Further details and analysis of FEWS can be found in our EACL21 paper, FEWS: Large-Scale, Low-Shot Word Sense Disambiguation with the Dictionary. If you use FEWS in your own work, please cite this paper.
Click here to download the FEWS dataset (87 MB).
You can find the code used to create FEWS from Wiktionary data in the Github repo here.
For any questions or comments about FEWS, please contact Terra Blevins at blvns@cs.washington.edu.