Create prepration.py

prepration.py

@@ -0,0 +1,20 @@
+from itertools import groupby
+from torch_snippets import *
+from datasets import load_dataset
+
+
+ds = load_dataset('darentang/sroie')
+classes = ds['train'].features['ner_tags'].feature.names
+
+def make_classes(X):
+    x = AD(X)
+    words_with_labels = wwl = [(w, classes[l]) for w,l in zip(x.words, x.ner_tags)]
+    wwl = [i for i in wwl if i[1] != 'O']
+    wwl = [(w,l.split('-')[1]) for w,l in wwl]
+    wwl = {cls: ' '.join(next(zip(*group))) for cls,group in groupby(wwl, key=lambda x: x[1])}
+    X['fields'] = wwl
+    return X
+    
+
+ds = ds.map(make_classes)
+ds.push_to_hub('sizhkhy/SROIE', commit_message="add fields as a new key")