Datasets:

cawoylel

/

OpenFulaSpeechCorpora

like 0

Hi,
Yes, actually this is raw data, we haven't applied any type of processing other than automatically aligning the text and audio into sentences. In the future it will be nice to clean this up somehow.
It's pretty trivial to edit this data and clean it up. In python, with the huggingface datasets, it could look like this for example:

from datasets import load_dataset

data = load_dataset("cawoylel/OpenFulaSpeechCorpora", "pulaar") # loading for the pulaar variety only

# cleaning pipeline
def clean(item):
    item["transcription"] = ...
    ...
    return item

# apply the cleaning
data = data.map(clean)

# push to hub
data.push_to_hub("your_repo")

If you want to help make the data better, we'd love to see your contribution.

We can also share with you the raw data (.wav and .txt files)

Hi Yaya,
Thank you for your quick reply.
Yes I wrote to you on cawoyel, I would like to contribute.
I am currently working on a fula translator and gathered resources that could also interest you for speech to text : https://github.com/flutter-painter/awesome_fula_nl_resources

Hello @flutter-painter ,
We sincerely apologize for not seeing your previous email. It's unfortunate that it ended up in our spam folder. I have sent you an invitation to join the slack channel.
Looking forward to connecting further and exploring the exciting possibilities ahead !