[Update] Update

custom_common_voice.py

@@ -17,6 +17,7 @@
 
 import datasets
 from datasets.tasks import AutomaticSpeechRecognition
+import pandas as pd
 
 
 _DATA_URL = "https://drive.google.com/uc?export=download&id=18ssh3gjpsMmjBfaIhcCwrfJMAvsdLGaI"
@@ -165,21 +166,18 @@ class CustomCommonVoice(datasets.GeneratorBasedBuilder):
 
         # audio is not a header of the csv files
         data_fields.remove("audio")
-        path_idx = data_fields.index("file_path")
-        script_idx = data_fields.index("script")
         examples = {}
 
-        with open(tsv_files, encoding="utf-8") as f:
-            lines = f.readlines()
-            for line in lines[1:]:
-                field_values = line.split("\t")
-                # set full path for mp3 audio file
-                audio_path = path_to_clips + "/" + field_values[path_idx]
-                script = field_values[script_idx]
-                examples[audio_path] = {
-                    "file_path": audio_path,
-                    "script": script,
-                }
+        df = pd.read_csv(tsv_files, sep="\t", header=0)
+        df = df.dropna()
+
+        for file_path, script in zip(df["file_path"], df["script"]):
+            # set full path for mp3 audio file
+            audio_path = path_to_clips + "/" + file_path
+            examples[audio_path] = {
+                "file_path": audio_path,
+                "script": script,
+            }
 
         inside_clips_dir = False
         for path, f in audio_files: