Create news-unfold-dataset.py

news-unfold-dataset.py

@@ -0,0 +1,47 @@
+from __future__ import absolute_import, division, print_function
+
+import datasets
+import os
+import pandas as pd
+from ast import literal_eval
+
+_DESCRIPTION = "dataset focused on media bias classification" 
+
+_URL = "news_unfold_new_sentences_final.csv"
+_HOMEPAGE = ""
+_LICENSE = ""
+
+class BABE(datasets.GeneratorBasedBuilder):
+  
+  def _info(self):
+    return datasets.DatasetInfo(
+        description=_DESCRIPTION,
+        citation=_CITATION,
+        features=datasets.Features(
+            {"text":datasets.Value("string"),
+            "news_link":datasets.Value("string"),
+            "outlet":datasets.Value("string"),
+            "topic":datasets.Value("string"),
+            "label":datasets.Value("int16")}
+        )
+    )
+    
+    
+    
+  def _split_generators(self, dl_manager):
+        path = dl_manager.download_and_extract(_URL)
+
+        # There is no predefined train/val/test split for this dataset.
+        return [
+            datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"file_path": path}),
+        ]
+  
+  def _generate_examples(self, file_path):
+       df = pd.read_csv(file_path)
+        
+       for i, row in df.iterrows():
+                yield i, {"text":row['text'],
+                          "label":row['label'],
+                          "news_link":str(row['news_link']),
+                          "outlet":str(row['outlet']),
+                          "topic":str(row['topic'])}