script

LongConL.py

@@ -1,63 +1,99 @@
 import datasets
-import os
+from datasets import DatasetInfo, Features, Value, Sequence, SplitGenerator
 import pandas as pd
+from pathlib import Path
+
+# Constants for the dataset
+_CITATION = """"""
+_DESCRIPTION = """Legal opinion classification dataset containing multiple tasks."""
+_HOMEPAGE = "https://huggingface.co/datasets/reglab/LongConL"
+
+# Define your configurations for each task dynamically
+TASKS = ["ATS-Jurisdiction"]  # Add all task names here
+
+# Configs for each task
+_CONFIGS = {
+    task: {
+        "description": f"Task: {task}",
+        "features": Features({
+            "Citation": Value("string"),
+            "Case Name": Value("string"),
+            "Opinion Text": Value("string"),
+            "Numerical Label": Value("string"),  # Assuming labels are strings, adjust if different
+            "Text Label": Value("string"),
+        }),
+        "license": None,
+    }
+    for task in TASKS
+}
+
 
 class LongConLDataset(datasets.GeneratorBasedBuilder):
-    """Legal opinion classification dataset for LongConL tasks."""
+    """Legal classification tasks for LongConL"""
 
-    # List of tasks (add more as needed)
+    # Define configurations for each task in the dataset
     BUILDER_CONFIGS = [
         datasets.BuilderConfig(
-            name=task_name, version=datasets.Version("1.0.0"), description=f"Task: {task_name}"
+            name=task, version=datasets.Version("1.0.0"), description=f"Task: {task}",
         )
-        for task_name in ["ATS-Jurisdiction", "ATS-FavorableJudgment"]  
+        for task in _CONFIGS
     ]
 
     def _info(self):
-        """Return dataset information."""
-        features = datasets.Features({
-            "Citation": datasets.Value("string"),
-            "Case Name": datasets.Value("string"),
-            "Opinion Text": datasets.Value("string"),
-            "Numerical Label": datasets.Value("string"),
-            "Text Label": datasets.Value("string"),
-        })
+        """Return dataset information with task-specific features."""
         return datasets.DatasetInfo(
-            description="Legal classification tasks dataset",
-            features=features,
-            homepage="https://huggingface.co/datasets/LongConL",
-            citation="",
+            description=_DESCRIPTION,
+            features=_CONFIGS[self.config.name]["features"],
+            homepage=_HOMEPAGE,
+            citation=_CITATION,
+            license=_CONFIGS[self.config.name]["license"],
         )
 
     def _split_generators(self, dl_manager):
-        """Split the dataset into train, validation, and test."""
-        task_name = self.config.name  # Dynamically get the current task name
-        base_dir = f"data/LongConL-tasks/{task_name}"
-        
+        """Return SplitGenerators based on the task name."""
+        task_name = self.config.name
+        base_dir = Path(f"data/LongConL-tasks/{task_name}")  # Path to task-specific directory
+
         return [
-            datasets.SplitGenerator(
+            SplitGenerator(
                 name=datasets.Split.TRAIN,
-                gen_kwargs={"file_path": f"{base_dir}/train.csv"},
+                gen_kwargs={"file_path": base_dir / "train.csv"},
             ),
-            datasets.SplitGenerator(
+            SplitGenerator(
                 name=datasets.Split.VALIDATION,
-                gen_kwargs={"file_path": f"{base_dir}/validation.csv"},
+                gen_kwargs={"file_path": base_dir / "validation.csv"},
             ),
-            datasets.SplitGenerator(
+            SplitGenerator(
                 name=datasets.Split.TEST,
-                gen_kwargs={"file_path": f"{base_dir}/test.csv"},
+                gen_kwargs={"file_path": base_dir / "test.csv"},
             ),
         ]
 
     def _generate_examples(self, file_path):
-        """Generate examples from a CSV file."""
+        """Yields examples from the CSV files."""
         data = pd.read_csv(file_path)
-        for id_, row in data.iterrows():
-            yield id_, {
+
+        # Iterate through each row and yield as example
+        for idx, row in data.iterrows():
+            yield idx, {
                 "Citation": row["Citation"],
                 "Case Name": row["Case Name"],
                 "Opinion Text": row["Opinion Text"],
-                "Numerical Label": row.get("Numerical Label", None),  # Optional column
+                "Numerical Label": row.get("Numerical Label", None),  # Optional column for tasks that may not have it
                 "Text Label": row["Text Label"],
             }
 
+
+# Example usage
+if __name__ == "__main__":
+    # Load a specific task dataset by task name
+    dataset = datasets.load_dataset("reglab/LongConL", name="ATS-Jurisdiction")
+    
+    # Access train, validation, and test splits
+    train_dataset = dataset["train"]
+    validation_dataset = dataset["validation"]
+    test_dataset = dataset["test"]
+
+    # Print some examples
+    print(train_dataset)
+