-
NewsMet.csv consists of the raw data. The candidates column consists of the top 200 candidates, literal_cands and metaphorical_cands consists of silver label candidates identified by the metaphor detection model, meaning_preserved_cands consists of the indentified gold label candidates -> if the sentence and verb are literal the candidates under this column would transform the sentence to metaphorical and vice versa.
-
custom_test_sets directory holds test set 1 and test set 2 as mentioned in the paper.
-
train_val_test_gold_plus directory consists of the standard splits. There is no overlap between the test set and other sets. Gold plus samples are identified by the sample_type column.
-
Label 1: metaphorical, Label 0: literal.
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|