Upload alphanumeric-audio-dataset.py

alphanumeric-audio-dataset.py

@@ -0,0 +1,75 @@
+import os
+import pandas as pd
+from datasets import Dataset, Audio
+
+def load_metadata(metadata_path):
+    """Load the metadata CSV file."""
+    return pd.read_csv(metadata_path)
+
+def generate_dataset_dict(metadata, audio_folder):
+    """
+    Generate the dataset dictionary by mapping metadata to corresponding audio files.
+    """
+    data = {
+        "file_name_id": [],
+        "audio": [],
+        "text": [],
+        "age": [],
+        "gender": [],
+        "nationality": [],
+        "native_language": [],
+        "familiarity_with_english": [],
+        "accent_strength": [],
+        "difficulties": [],
+        "recording_machine": [],
+    }
+    
+    for id in metadata['Response_ID'].unique():
+
+        df_subset = metadata[metadata['Response_ID']==id]
+        
+        # Collect audio paths for different categories
+        name_audio_path = df_subset['file_name'].iloc[0]
+        number_audio_path = df_subset['file_name'].iloc[2]
+        address_audio_path = df_subset['file_name'].iloc[1]
+        
+        row = df_subset.iloc[0]
+
+        # Only include data if all three audio files exist
+        if name_audio_path and number_audio_path and address_audio_path:
+            data["file_name_id"].append(id)
+            data["audio"].append({
+                "name_audio": name_audio_path,
+                "number_audio": number_audio_path,
+                "address_audio": address_audio_path,
+            })
+            data["text"].append({
+                "name": row["Name"],
+                "number": row["Number"],
+                "address": row["Address"],
+            })
+            data["age"].append(row["Age"])
+            data["gender"].append(row["Gender"])
+            data["nationality"].append(row["Nationality"])
+            data["native_language"].append(row["Native Language"])
+            data["familiarity_with_english"].append(row["Familiarity with English"])
+            data["accent_strength"].append(row["Accent Strength (Self reported)"])
+            data["difficulties"].append(row["Difficulties"])
+            data["recording_machine"].append(row["Recording Machine"])
+    
+    return data
+
+def load_dataset(metadata_path="metadata.csv", audio_folder="audio_data"):
+    """
+    Load the dataset, mapping metadata to audio and other fields.
+    """
+    metadata = load_metadata(metadata_path)
+    dataset_dict = generate_dataset_dict(metadata, audio_folder)
+    
+    # Use the Audio feature from the Hugging Face Datasets library
+    return Dataset.from_dict(dataset_dict).cast_column("audio", Audio())
+
+if __name__ == "__main__":
+    # Load the dataset and print a sample
+    dataset = load_dataset()
+    print(dataset)