---
dataset_info:
- config_name: all
  features:
  - name: id
    dtype: string
  - name: audio
    dtype: audio
  - name: audio_language
    dtype: string
  - name: text
    dtype: string
  - name: prompt
    dtype: string
  - name: duration
    dtype: float64
  - name: speaker_id
    dtype: string
  splits:
  - name: train
    num_bytes: 10017108673
    num_examples: 24840
  - name: dev
    num_bytes: 1098068265
    num_examples: 2535
  - name: dev_test
    num_bytes: 540536401
    num_examples: 1413
  download_size: 11646212136
  dataset_size: 11655713339
- config_name: new_duration_30s
  features:
  - name: id
    dtype: string
  - name: audio
    dtype: audio
  - name: audio_language
    dtype: string
  - name: text
    dtype: string
  - name: type
    dtype: string
  - name: prompt
    dtype: string
  - name: duration
    dtype: float64
  - name: speaker_id
    dtype: string
  splits:
  - name: train
    num_bytes: 6157169634
    num_examples: 116180
  - name: dev
    num_bytes: 662997924
    num_examples: 13825
  - name: dev_test
    num_bytes: 361871063
    num_examples: 6910
  download_size: 7153302446
  dataset_size: 7182038621
configs:
- config_name: all
  data_files:
  - split: train
    path: all/train-*
  - split: dev
    path: all/dev-*
  - split: dev_test
    path: all/dev_test-*
- config_name: new_duration_30s
  data_files:
  - split: train
    path: new_duration_30s/train-*
  - split: dev
    path: new_duration_30s/dev-*
  - split: dev_test
    path: new_duration_30s/dev_test-*
language:
- ki
---

# Kikuyu Speech Dataset

This dataset contains **Kikuyu speech audio with transcriptions**, sourced and adapted from the [ANV Data KE](https://huggingface.co/datasets/MCAA1-MSU/anv_data_ke) dataset.  
It includes both **scripted** and **unscripted** recordings, across different domains.  

We provide two configurations for different use cases.

---

## 📂 Dataset Structure

### Configurations
- **all**: Full dataset (all durations, **unscripted only**).  
- **new_duration_30s**: Audio samples ≤ 30 seconds (**scripted + unscripted**).  

### Features
Each entry has the following fields:

- `id`: string  
- `audio`: Audio()  
- `audio_language`: string  
- `text`: string (transcription if available)  
- `type`: string (`scripted` or `unscripted`)  
- `prompt`: string
- `duration`: float64 (in seconds)  
- `speaker_id`: string  

---

## 📊 Dataset Statistics

### **Unscripted only (all durations, `all`)**
| Split       | Hours   | #Examples |
|-------------|---------|-----------|
| train       | 322.2h  | 24,840    |
| dev         | 35.4h   | 2,535     |
| dev_test    | 17.2h   | 1,413     |

---


### **Scripted + Unscripted (≤30s, `new_duration_30s`)**

| Metric           | train    | dev     | dev_test | **Total** |
|------------------|---------:|--------:|---------:|----------:|
| Total Hours      | 185.8h   | 20.0h   | 10.9h    | 216.7h    |
| Scripted Hours   | 153.6h   | 16.6h   | 8.2h     | 178.4h    |
| Unscripted Hours | 32.2h    | 3.3h    | 2.6h     | 38.1h     |
| #Examples        | 116,180  | 13,825  | 6,910    | 136,915   |

---

## 🛠️ Usage

```python
from datasets import load_dataset

# Load full unscripted dataset
ds = load_dataset("evie-8/kikuyu-data", "all")

# Load 30s filtered dataset (scripted + unscripted)
ds_30s = load_dataset("evie-8/kikuyu-data", "new_duration_30s")

print(ds_30s["train"][0])