Update bucc-bitext-mining.py

bucc-bitext-mining.py

@@ -10,7 +10,7 @@ import gzip
 logger = datasets.logging.get_logger(__name__)
 
 _DESCRIPTION = """\
-        Tatoeba multilingual test set
+        BUCC 2018 Shared Task test dataset
 """
 
 _LANGUAGES = [
@@ -39,9 +39,9 @@ class BUCC(datasets.GeneratorBasedBuilder):
             description=_DESCRIPTION,
             features=datasets.Features(
                 {
-                    "sentence1": datasets.Value("string"),
-                    "sentence2": datasets.Value("string"),
-                    "gold": datasets.Sequence(datasets.Value("int32"), length=2)
+                    "sentence1": datasets.Sequence(datasets.Value("string")),
+                    "sentence2": datasets.Sequence(datasets.Value("string")),
+                    "gold": datasets.Sequence([datasets.Value("int32")])
                 },
             ),
             supervised_keys=None,
@@ -64,10 +64,5 @@ class BUCC(datasets.GeneratorBasedBuilder):
     def _generate_examples(self, text_path):
         """Yields examples."""
         with gzip.open(text_path,'rt') as f:
-            data = json.load(f)
-            for i, line in enumerate(zip(data['sentence1'], data['sentence2'], data['gold'])):
-                yield i, {
-                    'sentence1': line[0],
-                    'sentence2': line[1],
-                    'gold': [int(idx) for idx in line[2]],
-                }
+            for i, line in enumerate(f):
+                yield i, json.loads(line)