---
license: cc-by-sa-4.0
language:
  - tig
tags:
  - Tigre language
  - Eritrea
  - Audio
  - Speech Corpus
  - ASR
  - low-resource
pretty_name: Tigre Broadcast Speech Corpus
---

# 🇪🇷 Tigre Speech Corpus (Broadcast Audio)

A large-scale, open-source speech dataset for the **Tigre language** (ISO 639-3: `tig`), developed to support **Automatic Speech Recognition (ASR)**, **speech technology research**, and **language documentation** for one of the least-resourced languages in the Afro-Asiatic family.

This corpus provides **hundreds of hours of real-world spoken Tigre**, sourced from long-form public radio programming, making it one of the most substantial publicly available speech resources for any low-resource East African language.

---

## Dataset Summary

The dataset is derived from **public radio broadcasts**, covering a wide range of non-fiction spoken genres:

- News bulletins
- Long-form reports
- Radio interviews
- Documentaries
- Public-interest programming
  These recordings provide natural speech, varied speaking styles, and realistic acoustic conditions—ideal for training **robust ASR systems**.

---

## Key Statistics

| Feature            | Detail                                                    |
| ------------------ | --------------------------------------------------------- |
| **Language**       | Tigre (`tig`)                                             |
| **Source**         | Public Radio Broadcasts (News, Interviews, Documentaries) |
| **File Format**    | MP3                                                       |
| **Total Files**    | 2,227                                                     |
| **Total Duration** | **~582 hours**                                            |

This scale makes the corpus valuable for both **supervised ASR** and **self-supervised speech representation learning**.

---

## Duration & Distribution

The dataset contains **2,227 MP3 audio files** with a combined duration of **approximately 582 hours**.  
The files naturally fall into three nearly equal length categories:

- **Short:** 1–11 minutes
- **Medium:** 11–14 minutes
- **Long:** 14–61 minutes

This balanced distribution is especially beneficial for models that require **varied utterance lengths** and **diverse acoustic environments**.

---

## Intended Use

This dataset is ideal for:

### Speech Technology

1. Training and finetuning **ASR models** (e.g., Wav2Vec2.0, HuBERT, Whisper, MMS).
2. Pretraining **self-supervised speech encoders** for low-resource African languages.
3. Developing **acoustic models**, **segmentation systems**, and **forced aligners**.

### Research & Linguistics

4. Studying Tigre phonetics, prosody, and dialectal variation.
5. Supporting linguistic documentation and preservation efforts.
6. Building evaluation benchmarks for East African low-resource speech processing.

### Societal Impact

7. Creating digital tools for accessibility, education, and public information.
8. Expanding language technology inclusion for Eritrean languages.
9. Supporting nonprofit and humanitarian applications requiring speech resources.

---

## Ethical & Licensing Notes

- All recordings originate from **publicly available radio broadcasts**.
- The dataset is provided under **CC BY-SA 4.0**, allowing research and derivative works with attribution.
- Users must ensure responsible and culturally respectful use.
- The corpus must **not** be used for surveillance, profiling, or any harmful activity.

---

## Acknowledgments

This dataset is made possible through the efforts of:

- Tigre-speaking contributors and archivists preserving audio material
- The broader Eritrean community supporting language resource creation
- Researchers working to expand technology access for low-resource languages