Upload librispeech_asr_adversarial.py

librispeech_asr_adversarial.py

@@ -28,7 +28,8 @@ import datasets
 _CITATION = """\
 @inproceedings{panayotov2015librispeech,
   title={Librispeech: an ASR corpus based on public domain audio books},
-  author={Panayotov, Vassil and Chen, Guoguo and Povey, Daniel and Khudanpur, Sanjeev},
+  author={Panayotov, Vassil and Chen,
+      Guoguo and Povey, Daniel and Khudanpur, Sanjeev},
   booktitle={Acoustics, Speech and Signal Processing (ICASSP), 2015 IEEE International Conference on},
   pages={5206--5210},
   year={2015},
@@ -54,7 +55,13 @@ dataset = dataset.map(map_to_array, remove_columns=["file"])
 ```
 """
 # TODO: change
-_DL_URL = "https://drive.google.com/file/d/1oaBhaHlY4TD2JcvenR-6OZNIsyPG8OGN/view?usp=sharing"
+_DL_URL = "https://data.mendeley.com/api/datasets/kbdt923ybx/draft/files/65e7d2bf-16f7-4eed-8ac3-e3ae5d67aa40?a=7e4a4a33-35dd-4608-9d3a-24c0de170260"
+
+_DL_URLS = {
+    "adv": {
+        "test": _DL_URL,
+    }
+}
 
 
 class LibrispeechASRConfig(datasets.BuilderConfig):
@@ -97,43 +104,54 @@ class LibrispeechASR(datasets.GeneratorBasedBuilder):
         )
 
     def _split_generators(self, dl_manager):
+
         archive_path = dl_manager.download_and_extract(
-            _DL_URL)
+            _DL_URLS[self.config.name])
+
         return [
-            datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={
-                                    "archive_path": archive_path, "split_name": f"dev_{self.config.name}"}),
+            datasets.SplitGenerator(name="natural_nat_txt", gen_kwargs={
+                                    "archive_path": archive_path["test"], "split_name": f"natural_nat_txt"}),
+            datasets.SplitGenerator(name="adv_0.04_nat_txt", gen_kwargs={
+                                    "archive_path": archive_path["test"], "split_name": f"adv_0.04_nat_txt"}),
+            datasets.SplitGenerator(name="adv_0.015_nat_txt", gen_kwargs={
+                                    "archive_path": archive_path["test"], "split_name": f"adv_0.015_nat_txt"}),
+            datasets.SplitGenerator(name="adv_0.015_RIR_nat_txt", gen_kwargs={
+                                    "archive_path": archive_path["test"], "split_name": f"adv_0.015_RIR_nat_txt"}),
+            datasets.SplitGenerator(name="natural_adv_txt", gen_kwargs={
+                                    "archive_path": archive_path["test"], "split_name": f"natural_adv_txt"}),
+            datasets.SplitGenerator(name="adv_0.04_adv_txt", gen_kwargs={
+                                    "archive_path": archive_path["test"], "split_name": f"adv_0.04_adv_txt"}),
+            datasets.SplitGenerator(name="adv_0.015_adv_txt", gen_kwargs={
+                                    "archive_path": archive_path["test"], "split_name": f"adv_0.015_adv_txt"}),
+            datasets.SplitGenerator(name="adv_0.015_RIR_adv_txt", gen_kwargs={
+                                    "archive_path": archive_path["test"], "split_name": f"adv_0.015_RIR_adv_txt"}),
         ]
 
     def _generate_examples(self, archive_path, split_name):
         """Generate examples from a Librispeech archive_path."""
-        transcripts_glob = os.path.join(archive_path, "**/*.txt")
-        if split_name.endswith("adv-txt"):
-            split_folder = split_name[:-7]
+        transcript_file = os.path.join(archive_path, "manifest.txt")
+        if split_name.endswith("_adv_txt"):
+            split_folder = split_name[: -8]
             use_adv_transcript = True
         else:
-            assert split_name.endswith("nat-txt")
-            split_folder = split_name[:-7]
+            assert split_name.endswith("_nat_txt")
+            split_folder = split_name[: -8]
             use_adv_transcript = False
 
-        for transcript_file in glob.glob(transcripts_glob):
-            path = os.path.dirname(transcript_file)
-            audio_path = os.path.join(path, split_folder)
-            with open(os.path.join(path, transcript_file)) as f:
-                for line in f:
-                    line = line.strip()
-                    key, og_transcript, adv_transcript = line.split(",", 2)
-                    transcript = adv_transcript if use_adv_transcript else og_transcript
-                    suffix = "adv" if use_adv_transcript else "nat"
-                    audio_file = f"{key}_{suffix}.wav"
-                    speaker_id, chapter_id = [int(el)
-                                              for el in key.split("-")[:2]]
-                    split_key = key+"_"+suffix+"_"+split_name
-                    example = {
-                        "id": split_key,
-                        "speaker_id": speaker_id,
-                        "chapter_id": chapter_id,
-                        "file": os.path.join(audio_path, audio_file),
-                        "audio": os.path.join(audio_path, audio_file),
-                        "text": transcript,
-                    }
-                    yield split_key, example
+        path = os.path.dirname(transcript_file)
+        audio_path = os.path.join(path, split_folder)
+        with open(transcript_file) as f:
+            for line in f:
+                line = line.strip()
+                key, og_transcript, adv_transcript = line.split(",", 2)
+                transcript = adv_transcript if use_adv_transcript else og_transcript
+                suffix = "nat" if split_folder == "natural" else "adv"
+                audio_file = f"{key}_{suffix}.wav"
+                split_key = key+"_"+suffix+"_"+split_name
+                example = {
+                    "id": split_key,
+                    "file": os.path.join(audio_path, audio_file),
+                    "audio": os.path.join(audio_path, audio_file),
+                    "text": transcript,
+                }
+                yield split_key, example