is UFSAC dataset still available?
brucewlee opened this issue · 4 comments
The link provided in the repo is deprecated. Is the mentioned UFSAC dataset the same as Princeton Wordnet Gloss Corpus here? https://wordnetcode.princeton.edu/glosstag.shtml
The stats look a little different, which makes me question if your "With WNGC" setup in Table 2 is using the same data as the other papers that claim a similar setup.
You can find UFSAC here and download Version 2.1.
Compared to the one we used (downloaded from the deprecated link), this one has some POS tag differences (e.g., VBN v.s. VBD).
However, since both VBN and VBD will be eventually mapped to VERB in the code, using this one to train the model won't differ a lot from the model we provided.
Thank you for the quick response. But how did you perform ESR on the FEWS dataset, though? The dataset is not based on WordNet, so it would have been difficult to find hypernym relationships.
Hypernym is not used for ESR on the FEWS dataset. You can check the code for details.
@sonodasong Yes. Thank you for sharing the great work. Enjoyed reading the paper!