Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -1,6 +1,7 @@
|
|
1 |
---
|
2 |
task_categories:
|
3 |
- automatic-speech-recognition
|
|
|
4 |
language:
|
5 |
- en
|
6 |
- fr
|
@@ -8,14 +9,20 @@ language:
|
|
8 |
- es
|
9 |
tags:
|
10 |
- music
|
11 |
-
|
|
|
|
|
|
|
|
|
12 |
---
|
13 |
|
14 |
# JamALT: A Formatting-Aware Lyrics Transcription Benchmark
|
15 |
|
16 |
JamALT is a revision of the [JamendoLyrics](https://github.com/f90/jamendolyrics) dataset, adapted for use as an automatic lyrics transcription (ALT) benchmark.
|
17 |
|
18 |
-
The lyrics have been revised according to the newly compiled [annotation guidelines](GUIDELINES.md).
|
|
|
|
|
19 |
|
20 |
## Loading the data
|
21 |
|
@@ -30,4 +37,12 @@ for example, use `load_dataset("audioshake/jam-alt", "es")` to load only the Spa
|
|
30 |
Other arguments can be specified to control audio loading:
|
31 |
- `with_audio=False` to skip loading audio.
|
32 |
- `sampling_rate` and `mono=True` to control the sampling rate and number of channels.
|
33 |
-
- `decode_audio=False` to skip decoding the audio and just get paths to the MP3 files.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
task_categories:
|
3 |
- automatic-speech-recognition
|
4 |
+
- automatic-lyrics-transcription
|
5 |
language:
|
6 |
- en
|
7 |
- fr
|
|
|
9 |
- es
|
10 |
tags:
|
11 |
- music
|
12 |
+
- lyrics
|
13 |
+
- evaluation
|
14 |
+
- benchmark
|
15 |
+
- transcription
|
16 |
+
pretty_name: 'JamALT: A Formatting-Aware Lyrics Transcription Benchmark'
|
17 |
---
|
18 |
|
19 |
# JamALT: A Formatting-Aware Lyrics Transcription Benchmark
|
20 |
|
21 |
JamALT is a revision of the [JamendoLyrics](https://github.com/f90/jamendolyrics) dataset, adapted for use as an automatic lyrics transcription (ALT) benchmark.
|
22 |
|
23 |
+
The lyrics have been revised according to the newly compiled [annotation guidelines](GUIDELINES.md), which include rules about spelling, punctuation, and formatting.
|
24 |
+
|
25 |
+
See the [project website](https://audioshake.github.io/jam-alt/) for details.
|
26 |
|
27 |
## Loading the data
|
28 |
|
|
|
37 |
Other arguments can be specified to control audio loading:
|
38 |
- `with_audio=False` to skip loading audio.
|
39 |
- `sampling_rate` and `mono=True` to control the sampling rate and number of channels.
|
40 |
+
- `decode_audio=False` to skip decoding the audio and just get paths to the MP3 files.
|
41 |
+
|
42 |
+
## Running evaluation
|
43 |
+
|
44 |
+
Use the [alt-eval](https://github.com/audioshake/alt-eval) package for evaluation:
|
45 |
+
```python
|
46 |
+
from alt_eval import compute_metrics
|
47 |
+
compute_metrics(dataset["text"], hypotheses, languages=dataset["languages"])
|
48 |
+
```
|