up

common_voice_13_0.py

@@ -50,6 +50,10 @@ _TRANSCRIPT_URL = _BASE_URL + "transcript/{lang}/{split}.tsv"
 
 _N_SHARDS_URL = _BASE_URL + "n_shards.json"
 
+_WHISPER_TRANSCRIPT_URL = "https://huggingface.co/datasets/distil-whisper/common_voice_13_0/resolve/main/transcription_data/greedy_search/"
+
+_WHISPER_TRANSCRIPT_URLs = _WHISPER_TRANSCRIPT_URL + "/{split}-transcription.txt"
+
 
 class CommonVoiceConfig(datasets.BuilderConfig):
     """BuilderConfig for CommonVoice."""
@@ -117,6 +121,7 @@ class CommonVoice(datasets.GeneratorBasedBuilder):
                 "locale": datasets.Value("string"),
                 "segment": datasets.Value("string"),
                 "variant": datasets.Value("string"),
+                "whisper_transcript": datasets.Value("string"),
             }
         )
 
@@ -138,6 +143,7 @@ class CommonVoice(datasets.GeneratorBasedBuilder):
 
         audio_urls = {}
         splits = ("train", "dev", "test", "other", "invalidated")
+        splits_alt = ("train", "validation", "test")
         for split in splits:
             audio_urls[split] = [
                 _AUDIO_URL.format(lang=lang, split=split, shard_idx=i) for i in range(n_shards[lang][split])
@@ -148,13 +154,16 @@ class CommonVoice(datasets.GeneratorBasedBuilder):
         meta_urls = {split: _TRANSCRIPT_URL.format(lang=lang, split=split) for split in splits}
         meta_paths = dl_manager.download_and_extract(meta_urls)
 
+        transcription_urls = {split: _WHISPER_TRANSCRIPT_URLs.format(split=split) for split in splits_alt}
+        transcript_archive_path = dl_manager.download(transcription_urls)
+
         split_generators = []
         split_names = {
             "train": datasets.Split.TRAIN,
             "dev": datasets.Split.VALIDATION,
             "test": datasets.Split.TEST,
         }
-        for split in splits:
+        for split, split_alt in zip(splits, splits_alt):
             split_generators.append(
                 datasets.SplitGenerator(
                     name=split_names.get(split, split),
@@ -162,13 +171,14 @@ class CommonVoice(datasets.GeneratorBasedBuilder):
                         "local_extracted_archive_paths": local_extracted_archive_paths.get(split),
                         "archives": [dl_manager.iter_archive(path) for path in archive_paths.get(split)],
                         "meta_path": meta_paths[split],
+                        "whisper_transcript": transcript_archive_path[split_alt],
                     },
                 ),
             )
 
         return split_generators
 
-    def _generate_examples(self, local_extracted_archive_paths, archives, meta_path):
+    def _generate_examples(self, local_extracted_archive_paths, archives, meta_path, whisper_transcript):
         data_fields = list(self._info().features.keys())
         metadata = {}
         with open(meta_path, encoding="utf-8") as f:
@@ -186,6 +196,14 @@ class CommonVoice(datasets.GeneratorBasedBuilder):
                         row[field] = ""
                 metadata[row["path"]] = row
 
+        whisper_transcripts = []
+
+        with open(whisper_transcript, encoding="utf-8") as f:
+            for row in f:
+                whisper_transcripts.append(row.rstrip("\n"))
+
+        idx = 0
+
         for i, audio_archive in enumerate(archives):
             for filename, file in audio_archive:
                 _, filename = os.path.split(filename)
@@ -196,6 +214,8 @@ class CommonVoice(datasets.GeneratorBasedBuilder):
                     result["audio"] = {"path": path, "bytes": file.read()}
                     # set path to None if the audio file doesn't exist locally (i.e. in streaming mode)
                     result["path"] = path if local_extracted_archive_paths else filename
+                    result["whisper_transcript"] = whisper_transcripts[idx]
 
                     yield path, result
 
+                    idx += 1