Tobius commited on
Commit
da1e86d
1 Parent(s): ea8462b

Update teric_asr_lab.py

Browse files
Files changed (1) hide show
  1. teric_asr_lab.py +9 -8
teric_asr_lab.py CHANGED
@@ -28,7 +28,7 @@ from .release_stats import STATS
28
 
29
 
30
  _CITATION = """\
31
- @inproceedings{commonvoice:2020,
32
  author = {Bateesa, T. and Wairagala, EP. },
33
  title = {Teric Lab: A Massively-Multilingual Speech Corpus},
34
  }
@@ -43,7 +43,7 @@ _BASE_URL = "https://huggingface.co/datasets/Tobius/teric_asr_lab/resolve/main/"
43
 
44
  _AUDIO_URL = _BASE_URL + "audio/{lang}/{split}/{lang}_{split}_{shard_idx}.tar"
45
 
46
- _TRANSCRIPT_URL = _BASE_URL + "transcript/{lang}/{split}.csv"
47
 
48
  _N_SHARDS_URL = _BASE_URL + "n_shards.json"
49
 
@@ -102,7 +102,8 @@ class CommonVoice(datasets.GeneratorBasedBuilder):
102
  )
103
  features = datasets.Features(
104
  {
105
- "audio clip": datasets.features.Audio(sampling_rate=48_000),
 
106
  "sentence": datasets.Value("string"),
107
  "up_votes": datasets.Value("int64"),
108
  "down_votes": datasets.Value("int64"),
@@ -165,8 +166,8 @@ class CommonVoice(datasets.GeneratorBasedBuilder):
165
  with open(meta_path, encoding="utf-8") as f:
166
  reader = csv.DictReader(f, delimiter="\t", quoting=csv.QUOTE_NONE)
167
  for row in tqdm(reader, desc="Reading metadata..."):
168
- if not row["audio clip"].endswith(".wav"):
169
- row["audio clip"] += ".wav"
170
  # # accent -> accents in CV 8.0
171
  # if "accents" in row:
172
  # row["accent"] = row["accents"]
@@ -175,7 +176,7 @@ class CommonVoice(datasets.GeneratorBasedBuilder):
175
  for field in data_fields:
176
  if field not in row:
177
  row[field] = ""
178
- metadata[row["audio clip"]] = row
179
 
180
  for i, audio_archive in enumerate(archives):
181
  for path, file in audio_archive:
@@ -184,6 +185,6 @@ class CommonVoice(datasets.GeneratorBasedBuilder):
184
  result = dict(metadata[filename])
185
  # set the audio feature and the path to the extracted file
186
  path = os.path.join(local_extracted_archive_paths[i], path) if local_extracted_archive_paths else path
187
- result["audio clip"] = {"audio clip": path, "bytes": file.read()}
188
- result["audio clip"] = path
189
  yield path, result
 
28
 
29
 
30
  _CITATION = """\
31
+ @inproceedings{TericLabs:2023,
32
  author = {Bateesa, T. and Wairagala, EP. },
33
  title = {Teric Lab: A Massively-Multilingual Speech Corpus},
34
  }
 
43
 
44
  _AUDIO_URL = _BASE_URL + "audio/{lang}/{split}/{lang}_{split}_{shard_idx}.tar"
45
 
46
+ _TRANSCRIPT_URL = _BASE_URL + "transcript/{lang}/{split}.tsv"
47
 
48
  _N_SHARDS_URL = _BASE_URL + "n_shards.json"
49
 
 
102
  )
103
  features = datasets.Features(
104
  {
105
+ "path": datasets.Value("string"),
106
+ "audio": datasets.features.Audio(sampling_rate=48_000),
107
  "sentence": datasets.Value("string"),
108
  "up_votes": datasets.Value("int64"),
109
  "down_votes": datasets.Value("int64"),
 
166
  with open(meta_path, encoding="utf-8") as f:
167
  reader = csv.DictReader(f, delimiter="\t", quoting=csv.QUOTE_NONE)
168
  for row in tqdm(reader, desc="Reading metadata..."):
169
+ if not row["path"].endswith(".wav"):
170
+ row["path"] += ".wav"
171
  # # accent -> accents in CV 8.0
172
  # if "accents" in row:
173
  # row["accent"] = row["accents"]
 
176
  for field in data_fields:
177
  if field not in row:
178
  row[field] = ""
179
+ metadata[row["path"]] = row
180
 
181
  for i, audio_archive in enumerate(archives):
182
  for path, file in audio_archive:
 
185
  result = dict(metadata[filename])
186
  # set the audio feature and the path to the extracted file
187
  path = os.path.join(local_extracted_archive_paths[i], path) if local_extracted_archive_paths else path
188
+ result["audio"] = {"path": path, "bytes": file.read()}
189
+ result["path"] = path
190
  yield path, result