Datasets:
The suggested way to use same_topic senario do not cover all the authors in the test set
#3
by
cpuyyp
- opened
I follow the example in the summary as follows.
train_ds = load_dataset('guardian_authorship', name="cross_topic_1", split='train[:60%]+validation[:60%]+test[:60%]')
tests_ds = load_dataset('guardian_authorship', name="cross_topic_1", split='train[-40%:]+validation[-40%:]+test[-40%:]')
However, simply
len(tests_ds.to_pandas()['author'].unique())
gives 11. But there are 13 authors.