RaphaelOlivier commited on
Commit
ef8f85a
1 Parent(s): 50a01e6

Upload librispeech_asr_adversarial.py

Browse files
Files changed (1) hide show
  1. librispeech_asr_adversarial.py +50 -32
librispeech_asr_adversarial.py CHANGED
@@ -28,7 +28,8 @@ import datasets
28
  _CITATION = """\
29
  @inproceedings{panayotov2015librispeech,
30
  title={Librispeech: an ASR corpus based on public domain audio books},
31
- author={Panayotov, Vassil and Chen, Guoguo and Povey, Daniel and Khudanpur, Sanjeev},
 
32
  booktitle={Acoustics, Speech and Signal Processing (ICASSP), 2015 IEEE International Conference on},
33
  pages={5206--5210},
34
  year={2015},
@@ -54,7 +55,13 @@ dataset = dataset.map(map_to_array, remove_columns=["file"])
54
  ```
55
  """
56
  # TODO: change
57
- _DL_URL = "https://drive.google.com/file/d/1oaBhaHlY4TD2JcvenR-6OZNIsyPG8OGN/view?usp=sharing"
 
 
 
 
 
 
58
 
59
 
60
  class LibrispeechASRConfig(datasets.BuilderConfig):
@@ -97,43 +104,54 @@ class LibrispeechASR(datasets.GeneratorBasedBuilder):
97
  )
98
 
99
  def _split_generators(self, dl_manager):
 
100
  archive_path = dl_manager.download_and_extract(
101
- _DL_URL)
 
102
  return [
103
- datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={
104
- "archive_path": archive_path, "split_name": f"dev_{self.config.name}"}),
 
 
 
 
 
 
 
 
 
 
 
 
 
 
105
  ]
106
 
107
  def _generate_examples(self, archive_path, split_name):
108
  """Generate examples from a Librispeech archive_path."""
109
- transcripts_glob = os.path.join(archive_path, "**/*.txt")
110
- if split_name.endswith("adv-txt"):
111
- split_folder = split_name[:-7]
112
  use_adv_transcript = True
113
  else:
114
- assert split_name.endswith("nat-txt")
115
- split_folder = split_name[:-7]
116
  use_adv_transcript = False
117
 
118
- for transcript_file in glob.glob(transcripts_glob):
119
- path = os.path.dirname(transcript_file)
120
- audio_path = os.path.join(path, split_folder)
121
- with open(os.path.join(path, transcript_file)) as f:
122
- for line in f:
123
- line = line.strip()
124
- key, og_transcript, adv_transcript = line.split(",", 2)
125
- transcript = adv_transcript if use_adv_transcript else og_transcript
126
- suffix = "adv" if use_adv_transcript else "nat"
127
- audio_file = f"{key}_{suffix}.wav"
128
- speaker_id, chapter_id = [int(el)
129
- for el in key.split("-")[:2]]
130
- split_key = key+"_"+suffix+"_"+split_name
131
- example = {
132
- "id": split_key,
133
- "speaker_id": speaker_id,
134
- "chapter_id": chapter_id,
135
- "file": os.path.join(audio_path, audio_file),
136
- "audio": os.path.join(audio_path, audio_file),
137
- "text": transcript,
138
- }
139
- yield split_key, example
 
28
  _CITATION = """\
29
  @inproceedings{panayotov2015librispeech,
30
  title={Librispeech: an ASR corpus based on public domain audio books},
31
+ author={Panayotov, Vassil and Chen,
32
+ Guoguo and Povey, Daniel and Khudanpur, Sanjeev},
33
  booktitle={Acoustics, Speech and Signal Processing (ICASSP), 2015 IEEE International Conference on},
34
  pages={5206--5210},
35
  year={2015},
 
55
  ```
56
  """
57
  # TODO: change
58
+ _DL_URL = "https://data.mendeley.com/api/datasets/kbdt923ybx/draft/files/65e7d2bf-16f7-4eed-8ac3-e3ae5d67aa40?a=7e4a4a33-35dd-4608-9d3a-24c0de170260"
59
+
60
+ _DL_URLS = {
61
+ "adv": {
62
+ "test": _DL_URL,
63
+ }
64
+ }
65
 
66
 
67
  class LibrispeechASRConfig(datasets.BuilderConfig):
 
104
  )
105
 
106
  def _split_generators(self, dl_manager):
107
+
108
  archive_path = dl_manager.download_and_extract(
109
+ _DL_URLS[self.config.name])
110
+
111
  return [
112
+ datasets.SplitGenerator(name="natural_nat_txt", gen_kwargs={
113
+ "archive_path": archive_path["test"], "split_name": f"natural_nat_txt"}),
114
+ datasets.SplitGenerator(name="adv_0.04_nat_txt", gen_kwargs={
115
+ "archive_path": archive_path["test"], "split_name": f"adv_0.04_nat_txt"}),
116
+ datasets.SplitGenerator(name="adv_0.015_nat_txt", gen_kwargs={
117
+ "archive_path": archive_path["test"], "split_name": f"adv_0.015_nat_txt"}),
118
+ datasets.SplitGenerator(name="adv_0.015_RIR_nat_txt", gen_kwargs={
119
+ "archive_path": archive_path["test"], "split_name": f"adv_0.015_RIR_nat_txt"}),
120
+ datasets.SplitGenerator(name="natural_adv_txt", gen_kwargs={
121
+ "archive_path": archive_path["test"], "split_name": f"natural_adv_txt"}),
122
+ datasets.SplitGenerator(name="adv_0.04_adv_txt", gen_kwargs={
123
+ "archive_path": archive_path["test"], "split_name": f"adv_0.04_adv_txt"}),
124
+ datasets.SplitGenerator(name="adv_0.015_adv_txt", gen_kwargs={
125
+ "archive_path": archive_path["test"], "split_name": f"adv_0.015_adv_txt"}),
126
+ datasets.SplitGenerator(name="adv_0.015_RIR_adv_txt", gen_kwargs={
127
+ "archive_path": archive_path["test"], "split_name": f"adv_0.015_RIR_adv_txt"}),
128
  ]
129
 
130
  def _generate_examples(self, archive_path, split_name):
131
  """Generate examples from a Librispeech archive_path."""
132
+ transcript_file = os.path.join(archive_path, "manifest.txt")
133
+ if split_name.endswith("_adv_txt"):
134
+ split_folder = split_name[: -8]
135
  use_adv_transcript = True
136
  else:
137
+ assert split_name.endswith("_nat_txt")
138
+ split_folder = split_name[: -8]
139
  use_adv_transcript = False
140
 
141
+ path = os.path.dirname(transcript_file)
142
+ audio_path = os.path.join(path, split_folder)
143
+ with open(transcript_file) as f:
144
+ for line in f:
145
+ line = line.strip()
146
+ key, og_transcript, adv_transcript = line.split(",", 2)
147
+ transcript = adv_transcript if use_adv_transcript else og_transcript
148
+ suffix = "nat" if split_folder == "natural" else "adv"
149
+ audio_file = f"{key}_{suffix}.wav"
150
+ split_key = key+"_"+suffix+"_"+split_name
151
+ example = {
152
+ "id": split_key,
153
+ "file": os.path.join(audio_path, audio_file),
154
+ "audio": os.path.join(audio_path, audio_file),
155
+ "text": transcript,
156
+ }
157
+ yield split_key, example