---
language:
- pcm
license: cc-by-sa-4.0
task_categories:
- automatic-speech-recognition
tags:
- audio
- speech
- naija
- pidgin
dataset_info:
  features:
  - name: id
    dtype: string
  - name: speaker
    dtype: string
  - name: transcription
    dtype: string
  - name: audio
    dtype: audio
  - name: clip_start_ms
    dtype: int64
  - name: clip_end_ms
    dtype: int64
  - name: source_sent_id
    dtype: string
  - name: source_url
    dtype: string
size_categories:
- 1K<n<10K
---
# **Naija-ASR-Corpus v2.0 (NAC-v2.0)**

*A Foundational Automatic Speech Recognition Corpus for Nigerian Pidgin (Naija, PCM)*

---

## 📌 Dataset Summary

**Naija-ASR-Corpus (NAC-v2.0)** is a speech dataset derived from the **Universal
Dependencies Naija Spoken Corpus (UD_Naija-NSC)**.

The NAC Team processed the original long-form recordings by:
1. **Segmenting** the audio into sentence-level clips.
2. **Aligning** each clip to its transcript (`text_ortho`) from the CoNLL-U source.
3. **Tagging** each sample with a unique id and the original speaker code.

---

## 🧩 Schema

| Column           | Type   | Description                                                        |
|------------------|--------|----------------------------------------------------------------------|
| `id`             | string | Unique identifier for this sample (UUID4)                          |
| `speaker`        | string | Speaker code from the source corpus (`speaker_id` in UD_Naija-NSC) |
| `transcription`  | string | ASR transcript text                                                 |
| `audio`          | audio  | The audio clip, 16kHz mono                                          |
| `clip_start_ms`  | int64  | Clip start offset (ms) in the original long-form recording          |
| `clip_end_ms`    | int64  | Clip end offset (ms) in the original long-form recording            |
| `source_sent_id` | string | Sentence id from the UD_Naija-NSC source                            |
| `source_url`     | string | URL of the original long-form recording                             |

Rows: **5883**

---

## 🛠️ Data Source & Attribution

This dataset is derived from the **Universal Dependencies Naija Spoken Corpus
(UD_Naija-NSC)**. The original continuous recordings were split into
sentence-level clips and matched with their transcripts to build ASR-ready
audio/text pairs.

> This dataset contains content derived from *UD_Naija-NSC (Universal
> Dependencies Naija Spoken Corpus)*, licensed under **CC-BY-SA 4.0**.
> - **Original Authors:** UD_Naija-NSC contributors.
> - **Original License:** https://creativecommons.org/licenses/by-sa/4.0/

---

## 📜 Licensing

Since this dataset is a derivative work of UD_Naija-NSC (CC-BY-SA 4.0), it is
released under the **Creative Commons Attribution-ShareAlike 4.0
International (CC BY-SA 4.0)** license.

You are free to:
- **Share** — copy and redistribute the material in any medium or format.
- **Adapt** — remix, transform, and build upon the material for any purpose,
  even commercially.

Under the following terms:
- **Attribution** — give appropriate credit to UD_Naija-NSC and the NAC Team.
- **ShareAlike** — distribute your contributions under the same license as
  the original.

---

## 👥 Contributors

### NAC Team (Dataset Curation & Processing)
- **Team Lead:** Augustine, Silver, Timmy
- **Contributors:** Bryan, Ekene, Emmanuella, Shamsa

### Original Data Creators
- **UD_Naija-NSC Team** (Universal Dependencies)

---

## 🙏 Acknowledgements

This dataset builds on the same UD_Naija-NSC segmentation-and-alignment
pipeline used to produce [`timniel/Pidgin_ASR_Dataset_Combined`](https://huggingface.co/datasets/timniel/Pidgin_ASR_Dataset_Combined).
Credit and thanks go to **timniel** and the NAC Team for that earlier
combined dataset and the processing approach it established — this release
extends it with per-sample unique IDs, speaker attribution, clip timing,
and source provenance fields.

---

Pushed to: `voicedata/pidginData`