Add dataset info

echr.py

@@ -1,4 +1,5 @@
 import datasets
+from datasets import Value, Sequence
 
 _CITATION = """\
 @inproceedings{chalkidis-etal-2019-neural,
@@ -18,6 +19,10 @@ _CITATION = """\
 """
 
 _HOMEPAGE = "https://archive.org/details/ECHR-ACL2019"
+_DESCRIPTION = """\
+The ECHR Cases dataset is designed for experimentation of neural judgment prediction, as in the original 2019 ACL paper "Neural Legal Judgment Prediction in English".
+"""
+
 
 ARTICLES = {
     "2": "Right to life",
@@ -63,8 +68,53 @@ ARTICLES = {
     "P13-3": "Prohibition of reservations",
 }
 
+
 class Echr(datasets.GeneratorBasedBuilder):
+    """ECHR dataset."""
+
     BUILDER_CONFIGS = [
         datasets.BuilderConfig(name="non-anon", data_dir="data"),
         datasets.BuilderConfig(name="anon", data_dir="data_anon"),
     ]
+
+    def _info():
+        features = {
+            "itemid": Value(dtype="string", id=None),
+            "languageisocode": Value(dtype="string", id=None),
+            "respondent": Value(dtype="string", id=None),
+            "branch": Value(dtype="string", id=None),
+            "date": Value(dtype="int64", id=None),
+            "docname": Value(dtype="string", id=None),
+            "importance": Value(dtype="int64", id=None),
+            "conclusion": Value(dtype="string", id=None),
+            "judges": Value(dtype="string", id=None),
+            "text": Sequence(
+                feature=Value(dtype="string", id=None), length=-1, id=None
+            ),
+            "violated_articles": Sequence(
+                feature=Value(dtype="string", id=None), length=-1, id=None
+            ),
+            "violated_paragraphs": Sequence(
+                feature=Value(dtype="string", id=None), length=-1, id=None
+            ),
+            "violated_bulletpoints": Sequence(
+                feature=Value(dtype="string", id=None), length=-1, id=None
+            ),
+            "non_violated_articles": Sequence(
+                feature=Value(dtype="string", id=None), length=-1, id=None
+            ),
+            "non_violated_paragraphs": Sequence(
+                feature=Value(dtype="string", id=None), length=-1, id=None
+            ),
+            "non_violated_bulletpoints": Sequence(
+                feature=Value(dtype="string", id=None), length=-1, id=None
+            ),
+            "violated": Value(dtype="bool", id=None),
+        }
+
+        return datasets.DatasetInfo(
+            features=features,
+            homepage=_HOMEPAGE,
+            description=_DESCRIPTION,
+            citation=_CITATION,
+        )