suru8-ai/chichewa_english_code_switch_dataset

Chichewa-English Code-Switched Speech Dataset

Dataset Description

A speech dataset containing 247 audio recordings of Chichewa-English code-switched phrases. Code-switching — the practice of alternating between two or more languages within a single conversation — is extremely common in Malawi and across multilingual African communities. This dataset captures that natural linguistic behavior in spoken form.

Purpose

This dataset is designed to support research and development in:

• Automatic Speech Recognition (ASR) for code-switched Chichewa-English speech
• Multilingual and low-resource NLP for African languages
• Code-switching detection and modeling

Dataset Structure

Files

• metadata.csv — contains transcriptions, speaker info, and audio file references
• audios/ — directory containing 247 .wav audio files

Metadata Fields

Field	Description
phrase	The transcribed text of the audio (code-switched)
speaker	Speaker identifier
file_name	Path to the corresponding .wav audio file

Example Entries

phrase	speaker	file_name
Moni brother, uli bwanji	Timothy	audios/1.wav
Good morning m'bale wanga, wazuka bwanji	Timothy	audios/3.wav
Ndikufuna ndikagule bread	Timothy	audios/41.wav
Phone yanga ili battery low	Timothy	audios/65.wav

Languages

• Chichewa (Nyanja) — a Bantu language widely spoken in Malawi, Zambia, and Mozambique
• English — used as a secondary language in education and formal settings in Malawi

Most utterances contain a mix of both languages within a single phrase, reflecting natural code-switching patterns.

Dataset Statistics

• Total recordings: 247
• Total size: ~74.8 MB
• Audio format: WAV
• Speaker(s): 1 (Timothy)
• Domain: Everyday conversational phrases (greetings, transport, shopping, health, technology, family, etc.)

Topics Covered

The phrases span a wide range of everyday scenarios:

• Greetings and farewells
• Transportation and travel
• Shopping and food
• Phone, internet, and technology
• Work and school
• Health
• Family
• Money and payments
• Time expressions

Usage

from datasets import load_dataset

dataset = load_dataset("suru8-ai/chichewa_english_code_switch_dataset")

Citation

If you use this dataset, please cite:

@dataset{chichewa_english_code_switch_2025,
  title={Chichewa-English Code-Switched Speech Dataset},
  author={Rexplore Research Labs},
  year={2026},
  url={https://huggingface.co/datasets/suru8-ai/chichewa_english_code_switch_dataset}
}

License

This dataset is released under the Apache 2.0 License.