Add the ReazonSpeech recipe #1330
Conversation
I created this recipe by copying "aishell4" recipe, and stripping the most of the contents. Signed-off-by: Fujimoto Seiji <fujimoto@ceptord.net>
…/lhotse Experiemntal version for ReazonSpeech
lhotse/recipes/reazonspeech.py
Outdated
| duration=item["duration"], | ||
| channel=0, | ||
| language="Japanese", | ||
| speaker=str(idx), |
There was a problem hiding this comment.
If speaker information is not available, just omit this field.
| from lhotse.utils import Pathlike | ||
|
|
||
|
|
||
| def prepare_reazonspeech( |
There was a problem hiding this comment.
Since reazonspeech is a large dataset this method may require a lot of cpu memory before it writes anything. Would you consider modifying the recipe to resemble gigaspeech more closely? It writes examples as it processes them for reduced memory usage https://github.com/lhotse-speech/lhotse/blob/master/lhotse/recipes/gigaspeech.py
|
Also, could you add an entry in the dataset table in docs/corpus.rst? |
|
Thanks for your quick feedback and suggestions! I will get back to your comments soon. :) |
|
Thanks for your quick feedback! I'll fix those checks very soon. :) |
|
It seems the tests are failing on importing |
Head branch was pushed to by a user without write access
|
@pzelasko Thanks for the note! I've already changed it to local import. Please feel free to let me know if there's anything else I need to improve. :) |
ReazonSpeech is an open-source dataset that contains a diverse set of natural Japanese speech, collected from terrestrial television streams. It contains more than 35,000 hours of audio.
The dataset is available on Hugging Face. For more details, please visit: