Datasets:

Modalities:
Audio
Text
Formats:
parquet
Size:
< 1K
ArXiv:
DOI:
Libraries:
Datasets
pandas
cifkao commited on
Commit
47b0aa0
1 Parent(s): 16a3c1f
README.md CHANGED
@@ -16,6 +16,119 @@ tags:
16
  - transcription
17
  pretty_name: 'JamALT: A Readability-Aware Lyrics Transcription Benchmark'
18
  paperswithcode_id: jam-alt
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
19
  ---
20
 
21
  # JamALT: A Readability-Aware Lyrics Transcription Benchmark
@@ -42,17 +155,18 @@ See the [project website](https://audioshake.github.io/jam-alt/) for details.
42
 
43
  ```python
44
  from datasets import load_dataset
45
- dataset = load_dataset("audioshake/jam-alt")["test"]
46
  ```
47
 
48
  A subset is defined for each language (`en`, `fr`, `de`, `es`);
49
  for example, use `load_dataset("audioshake/jam-alt", "es")` to load only the Spanish songs.
50
 
51
- By default, the dataset comes with audio. To skip loading the audio, use `with_audio=False`.
52
  To control how the audio is decoded, cast the `audio` column using `dataset.cast_column("audio", datasets.Audio(...))`.
53
  Useful arguments to `datasets.Audio()` are:
54
  - `sampling_rate` and `mono=True` to control the sampling rate and number of channels.
55
- - `decode=False` to skip decoding the audio and just get the MP3 file paths.
 
 
56
 
57
  ## Running the benchmark
58
 
@@ -61,14 +175,14 @@ The evaluation is implemented in our [`alt-eval` package](https://github.com/aud
61
  from datasets import load_dataset
62
  from alt_eval import compute_metrics
63
 
64
- dataset = load_dataset("audioshake/jam-alt", revision="v1.0.0")["test"]
65
  # transcriptions: list[str]
66
  compute_metrics(dataset["text"], transcriptions, languages=dataset["language"])
67
  ```
68
 
69
  For example, the following code can be used to evaluate Whisper:
70
  ```python
71
- dataset = load_dataset("audioshake/jam-alt", revision="v1.0.0")["test"]
72
  dataset = dataset.cast_column("audio", datasets.Audio(decode=False)) # Get the raw audio file, let Whisper decode it
73
 
74
  model = whisper.load_model("tiny")
@@ -78,9 +192,9 @@ transcriptions = [
78
  ]
79
  compute_metrics(dataset["text"], transcriptions, languages=dataset["language"])
80
  ```
81
- Alternatively, if you already have transcriptions, you might prefer to skip loading the audio:
82
  ```python
83
- dataset = load_dataset("audioshake/jam-alt", revision="v1.0.0", with_audio=False)["test"]
84
  ```
85
 
86
  ## Citation
@@ -108,4 +222,4 @@ When using the benchmark, please cite [our paper](https://www.arxiv.org/abs/2408
108
  address={Rhodes Island, Greece},
109
  doi={10.1109/ICASSP49357.2023.10096725}
110
  }
111
- ```
 
16
  - transcription
17
  pretty_name: 'JamALT: A Readability-Aware Lyrics Transcription Benchmark'
18
  paperswithcode_id: jam-alt
19
+ dataset_info:
20
+ - config_name: all
21
+ features:
22
+ - name: name
23
+ dtype: string
24
+ - name: text
25
+ dtype: string
26
+ - name: language
27
+ dtype: string
28
+ - name: license_type
29
+ dtype: string
30
+ - name: audio
31
+ dtype: audio
32
+ splits:
33
+ - name: test
34
+ num_bytes: 409411912.0
35
+ num_examples: 79
36
+ download_size: 409150043
37
+ dataset_size: 409411912.0
38
+ - config_name: de
39
+ features:
40
+ - name: name
41
+ dtype: string
42
+ - name: text
43
+ dtype: string
44
+ - name: language
45
+ dtype: string
46
+ - name: license_type
47
+ dtype: string
48
+ - name: audio
49
+ dtype: audio
50
+ splits:
51
+ - name: test
52
+ num_bytes: 107962802.0
53
+ num_examples: 20
54
+ download_size: 107942102
55
+ dataset_size: 107962802.0
56
+ - config_name: en
57
+ features:
58
+ - name: name
59
+ dtype: string
60
+ - name: text
61
+ dtype: string
62
+ - name: language
63
+ dtype: string
64
+ - name: license_type
65
+ dtype: string
66
+ - name: audio
67
+ dtype: audio
68
+ splits:
69
+ - name: test
70
+ num_bytes: 105135091.0
71
+ num_examples: 20
72
+ download_size: 105041371
73
+ dataset_size: 105135091.0
74
+ - config_name: es
75
+ features:
76
+ - name: name
77
+ dtype: string
78
+ - name: text
79
+ dtype: string
80
+ - name: language
81
+ dtype: string
82
+ - name: license_type
83
+ dtype: string
84
+ - name: audio
85
+ dtype: audio
86
+ splits:
87
+ - name: test
88
+ num_bytes: 105024257.0
89
+ num_examples: 20
90
+ download_size: 104979012
91
+ dataset_size: 105024257.0
92
+ - config_name: fr
93
+ features:
94
+ - name: name
95
+ dtype: string
96
+ - name: text
97
+ dtype: string
98
+ - name: language
99
+ dtype: string
100
+ - name: license_type
101
+ dtype: string
102
+ - name: audio
103
+ dtype: audio
104
+ splits:
105
+ - name: test
106
+ num_bytes: 91289764.0
107
+ num_examples: 19
108
+ download_size: 91218543
109
+ dataset_size: 91289764.0
110
+ configs:
111
+ - config_name: all
112
+ data_files:
113
+ - split: test
114
+ path: parquet/all/test-*
115
+ default: true
116
+ - config_name: de
117
+ data_files:
118
+ - split: test
119
+ path: parquet/de/test-*
120
+ - config_name: en
121
+ data_files:
122
+ - split: test
123
+ path: parquet/en/test-*
124
+ - config_name: es
125
+ data_files:
126
+ - split: test
127
+ path: parquet/es/test-*
128
+ - config_name: fr
129
+ data_files:
130
+ - split: test
131
+ path: parquet/fr/test-*
132
  ---
133
 
134
  # JamALT: A Readability-Aware Lyrics Transcription Benchmark
 
155
 
156
  ```python
157
  from datasets import load_dataset
158
+ dataset = load_dataset("audioshake/jam-alt", split="test")
159
  ```
160
 
161
  A subset is defined for each language (`en`, `fr`, `de`, `es`);
162
  for example, use `load_dataset("audioshake/jam-alt", "es")` to load only the Spanish songs.
163
 
 
164
  To control how the audio is decoded, cast the `audio` column using `dataset.cast_column("audio", datasets.Audio(...))`.
165
  Useful arguments to `datasets.Audio()` are:
166
  - `sampling_rate` and `mono=True` to control the sampling rate and number of channels.
167
+ - `decode=False` to skip decoding the audio and just get the MP3 file paths and contents.
168
+
169
+ The `load_dataset` function also accepts a `columns` parameter, which can be useful for example if you want to skip downloading the audio (see the example below).
170
 
171
  ## Running the benchmark
172
 
 
175
  from datasets import load_dataset
176
  from alt_eval import compute_metrics
177
 
178
+ dataset = load_dataset("audioshake/jam-alt", revision="v1.1.0", split="test")
179
  # transcriptions: list[str]
180
  compute_metrics(dataset["text"], transcriptions, languages=dataset["language"])
181
  ```
182
 
183
  For example, the following code can be used to evaluate Whisper:
184
  ```python
185
+ dataset = load_dataset("audioshake/jam-alt", revision="v1.1.0", split="test")
186
  dataset = dataset.cast_column("audio", datasets.Audio(decode=False)) # Get the raw audio file, let Whisper decode it
187
 
188
  model = whisper.load_model("tiny")
 
192
  ]
193
  compute_metrics(dataset["text"], transcriptions, languages=dataset["language"])
194
  ```
195
+ Alternatively, if you already have transcriptions, you might prefer to skip loading the `audio` column:
196
  ```python
197
+ dataset = load_dataset("audioshake/jam-alt", revision="v1.1.0", split="test", columns=["name", "text", "language", "license_type"])
198
  ```
199
 
200
  ## Citation
 
222
  address={Rhodes Island, Greece},
223
  doi={10.1109/ICASSP49357.2023.10096725}
224
  }
225
+ ```
jam-alt.py → loader.py RENAMED
@@ -11,7 +11,7 @@ from typing import Optional
11
  import datasets
12
 
13
 
14
- _VERSION = "1.0.0"
15
 
16
 
17
  _CITATION = """\
 
11
  import datasets
12
 
13
 
14
+ _VERSION = "1.1.0"
15
 
16
 
17
  _CITATION = """\
parquet/all/test-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ed027d2032652b082c79b63008eecea1dacce97400ce1d5cb0326e958302d6e1
3
+ size 409150043
parquet/de/test-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d4fe32692f20daeb06b2b3c253b51dbd5bc4a4c6bb491a64fde821d263d95134
3
+ size 107942102
parquet/en/test-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6c8870334ca9ff4a5166940355169ee36a260725c782a23bd352948d38c83f70
3
+ size 105041371
parquet/es/test-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96468681e62d4f69729c2d33890bfc17caf9718b4f56cd843ab1468a4c72c8a1
3
+ size 104979012
parquet/fr/test-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:935eb807ce731d44dfc904a1b8f2dd751b626e8c40774d56ee793b799ec987e0
3
+ size 91218543
push_to_hub.py ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import argparse
2
+ import datasets
3
+
4
+
5
+ def main():
6
+ parser = argparse.ArgumentParser()
7
+ parser.add_argument("--repo", type=str, required=True)
8
+ parser.add_argument("--revision", type=str, required=True)
9
+ args = parser.parse_args()
10
+
11
+ for config_name in ["all", "en", "es", "de", "fr"]:
12
+ dataset = datasets.load_dataset(
13
+ "./loader.py", config_name, trust_remote_code=True
14
+ )
15
+ dataset.push_to_hub(
16
+ args.repo,
17
+ config_name,
18
+ set_default=(config_name == "all"),
19
+ data_dir=f"parquet/{config_name}",
20
+ commit_message=f"Upload parquet/{config_name}",
21
+ revision=args.revision,
22
+ )
23
+
24
+
25
+ if __name__ == "__main__":
26
+ main()