Hugging Face's logo

Datasets:
ghananlpcommunity
/
ghana-english-speech-600hrs

Tasks:
Automatic Speech Recognition
Modalities:
Audio
Text
Formats:
parquet
Languages:
English
Size:
100K - 1M
Tags:
audio
speech
asr
ghanaian-english
west-african-english
Libraries:
Datasets
Dask
Polars
License:
Dataset card Data Studio Files
xet
 Community
Dataset Viewer
Auto-converted to Parquet Duplicate
audio
audioduration (s)
0.49
12
corrected_text
stringlengths
3
271
duration_ss
float32
2
12
He made mention that if your pastor tells you, if you ask your pastor who you should vote for and he mentions uh the name that you should vote for, that is the person you should vote for. Who
8.96
Dr. Yun. There is no two ways. I
2.08
Yeah, definitely. I'm a Methodist. I'm
2.72
a Methodist and I believe in my hymns.
3.76
Do you know this song? A child to keep I
2.08
have and a God to glorify. a time to
2.72
And that is what uh the Good Bible tells us. Look for knowledge. When you get the
4.96
knowledge, use it for the benefit of your people. He's a man that we can
3.84
rally support behind to ensure that we bring prosperity to the good people of Ghana.
6.01
brother, a friend, a confidant. I mean, we did things
6.64
together. I mean, uh the strategies that
2.96
we use for the election for money was three, four, five.
5.44
It's no more. So, we have to take the
2.08
pieces and put it together. and then uh
3.12
take on and and make sure that his legacy uh lives on. I mean that's the
5.12
only thing we have a huge legacy especially the the activate the body camera for the election directorate.
8.88
We will continue to to use it. Um it's a
4.4
code I mean he gave us the the tools to decode it and then it helped us a lot. So we will
6.64
designing for the party.
2.08
When you say that his legacy would continue to live on aside what you just mentioned, is there any other thing that all of you collectively coming together would do to make sure that indeed his legacy does live on?
12
Well, that's why we we said the principle on which he was living. Uh
5.36
I normally usually call him my my teacher and lecturer.
4.88
touching it it's hit us hard it hit us that and that is why we are going in the to Koforidua because he was passionate about it. That very morning
11.28
uh he he gave us a strategy. We had a
2.48
will the August 7 in the morning. By
3.28
7:45 we are done.
2
He didn't let us know that he was traveling. We took out the strategy.
3.76
We're going to start the implementation in the evening and report back to him only for him to lose his life in the afternoon. So you can imagine such a
8.24
person who was passionate in winning that bi-election it to be in the right place for us to go and win the election in his glory. I
9.2
mean to to to let him know that what he wanted we have actually achieved it for him.
5.84
would be a level of some level of disappointment to the memory of Dr. Manbama. Not certainly of course already
7.52
process before the the crash came. So we have to
4.08
continue to just show that the strategy he gave us we will execute it and make sure that it comes to pass.
6.56
What would be the message that you would give to the bereaved family?
2.96
They should take heart. They've lost a
2.24
son but we are there to represent them.
2.72
They they still have other sons that they can rely on. uh we will still give
5.28
them the shoulder they needed as a family. Uh yesterday I was with the
4.72
mother and then they listen and we are there for them. We will continue
5.2
whatever money started and and see to it logical conclusion we will continue it we are still there for the family. We
8.48
their life.
2.08
plowing.
2.32
Next slide please.
3.64
So this is a machine operating or doing the plowing. You have waffle planting
5.44
wing.
3
Next slide.
3.24
Waffle harvesting w receiver. And the
3.12
radio receiver it helps to provide information or indicate the power when the power is in the fire.
8.04
Then we have our four tracking body sensors.
2.32
And the factors include the blood pressure and the rate of the farmer. So
4.8
benefit of to the farmer is important for the farm but is also available to practice. It helps
10.56
to talk to the farmer before they start also helps to educate the farmer and the farmer and inform the farmer when the farmer is in the farm. W also help to
11.12
check the the body strength. They remove the environment
5.28
and social impact of environment to reduce the reliance on coal. Hence
6.56
reducing the impact on climate change.
2.16
official retired who is also with us. Uh
2.48
he obviously has followed all that happened. Thankfully a lot of this was
3.68
doesn't recur. Um, based on your
2.96
assessment, we had a lot of time now to look at the facts as we know it. Based
3.76
on your assessment, on your experience in the Ghana armed forces, what's your what's your own reading of the way this incident in the immediate aftermath was handled? One, securing the
11.04
site, two, um, the search and rescue efforts, all which would have fed into the quality investigations that you expect um, after an incident like that.
10.32
First, good evening to your good self, to my senior friend, wing commander, and my very learned friend Bright.
11.04
Let me express once again sincere condolences to all of us in the bereaved um family.
10.72
Now, I might have indicated already on news file that I I'm writing a paper.
7.44
What is clear to me is that 14 minutes after the helicopter took off at 09twelve it went off radar.
11.76
But wing commander will explain to us technically that when a plane is off radar, it does not mean that it is lost.
8.56
It is just that the radar cannot find it. So in this case, maybe the Z9 was
7.84
flying below 1,000 ft there about based on what raiders were operating um at any material time.
11.28
It should have taken the helicopter twelve sorry fifteen minutes all things being well to have arrived at Abuasi.
9.84
expecting this helicoptercopter to have landed at Abuasi by so and so time. Do you have any wind
8.48
of it? Can you if you like look for it?
4.8
stadium.
2.16
were in the forest has seen it. They
2.64
started moving to the crash site and has started taking away um broken parts and so which are critical exhibits. So in
10.08
day you and I were here when the coffins arrived in the night of the six.
4.8
nan Unless maybe I'm confused again.
5.6
which it did.
2.72
Then another plane fixed wing the Kasa twelve5 transport plane left ARA with investigation team. Now that Kasa twelve5
10.72
which sent investigation team on the same 6 of August was the one that returned with the coffins.
7.84
However, given the time that it took from 14 in fact as I said the military got to the scene 1400 hours. The regional minister
10.56
got to the scene at about fifteen00 hours with the fire service, with the police service, national ambulance service.
8
But there were largely body parts that were being pulled from call it the slope or the ditch. It took time. The original
9.52
minister had also intended that if it was possible they should find the black box.
7.2
Now that was when nightfall caught up with them. So therefore when it became
6.16
obvious that the site had been found, it was no longer necessary for the MI7 to fly into the air.
9.52
that after that you sanitize the ambulances. So that was what was done.
5.6
was waiting to now take the bodies transfer them from the sacks into proper body bags and then into the aircraft.
9.6
Indeed, the coffins that we saw in Ara were all arranged during the flight. The
6.88
bodies were put in the coffins on the flight. The Ghana flags were draped or
6.16
the coffins were draped with the Ghana flags during the flight. So these are
4.88
some of the circumstances.
2.16
It is not very accurate that the search team didn't get there in good enough time. Indeed, we are thankful
9.68
and pieces were falling in place. Now
3.68
talk about body bags. There were body
2.64
nan Yeah. I mean before I go to body bags on
2.08
the question of holding off the crash site and securing it. Um what what what
5.84
have you learned? Because yes team got
2.16
there and when we got there indeed there were military officers, there was Nadmo, there was police, there's fire service.
4.48
Yes. But people still had access to the
2.24
site. I mean overnight into the morning.
2.16
End of preview. Expand in Data Studio

🇬🇭 Ghana English ASR Dataset

A speech dataset of Ghanaian English extracted from Ghanaian news media broadcasts, designed for training and fine-tuning Automatic Speech Recognition (ASR) models on West African English accents.

📂 Dataset Structure

Column Type Description
audio Audio 24 kHz mono WAV audio segment
corrected_text string Verbatim transcription of the audio segment
duration_ss float Duration of the audio segment in seconds

📊 Statistics

Metric Value
Total clips 406,094
Total duration 619.52 hours
Mean clip duration 5.49 s
Min / Max clip duration 2.0 s / 12.0 s
Mean words per clip 14.3
Min / Max words 1 / 98
Vocabulary size 84,832 unique words
Sample rate 24,000 Hz (mono)

🚀 Usage 

from datasets import load_dataset

dataset = load_dataset("ghananlpcommunity/ghana-english-speech-600hrs")
train = dataset["train"]

example = train[0]
print("Transcription:", example["corrected_text"])
print("Duration (s):", example["duration_ss"])
print("Audio array shape:", example["audio"]["array"].shape)
print("Sample rate:", example["audio"]["sampling_rate"])

Fine-tuning with Whisper 

from transformers import WhisperProcessor

processor = WhisperProcessor.from_pretrained("openai/whisper-small")

def prepare_batch(batch):
    audio = batch["audio"]
    batch["input_features"] = processor(
        audio["array"], sampling_rate=audio["sampling_rate"], return_tensors="pt"
    ).input_features[0]
    batch["labels"] = processor.tokenizer(batch["corrected_text"]).input_ids
    return batch

dataset = dataset.map(prepare_batch, remove_columns=dataset.column_names)

🎯 Intended Use Cases

  • Fine-tuning Whisper, Wav2Vec2, MMS for Ghanaian / West African English
  • Building accent-aware ASR pipelines for Ghanaian broadcast media
  • Linguistic research on Ghanaian English phonology and prosody
  • Low-resource African language / dialect ASR benchmarking

⚠️ Limitations

  • Domain-specific: broadcast news only, may not generalise to conversational English.
  • Speaker diversity not formally audited.
  • Transcriptions may contain occasional errors in proper nouns.

📜 Citation 

@dataset{ghana_english_asr,
  author    = {Owusu, Mich-Seth},
  title     = {Ghana English ASR Dataset},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/ghananlpcommunity/ghana-english-speech-600hrs}
}

🙏 Acknowledgments

Created by Mich-Seth Owusu for the Ghana NLP Community.
Downloads last month
558
Number of rows:
406,094
Total file size:
105 GB