WARNING : This dataset has data leakage !
#7
by
mnaguib
- opened
Minimal example :
from datasets import load_dataset
train_dataset = load_dataset("Jean-Baptiste/wikiner_fr", split='train')
test_dataset = load_dataset("Jean-Baptiste/wikiner_fr", split='test')
train_sentences = [' '.join(w) for w in train_dataset['tokens']]
test_sentences = [' '.join(w) for w in test_dataset['tokens']]
leakage = set(train_sentences).intersection(set(test_sentences))
print("LEAKAGE BETWEEN TRAIN AND TEST SETS")
print(len(leakage)) #prints 23
In addition, data seems to have intra-split duplicates :
print("duplicate sentences in train dataset : ", len(train_sentences) - len(set(train_sentences))) #print 599
print("duplicate sentences in test dataset : ", len(test_sentences) - len(set(test_sentences))) #print 17