Lahja-IT: 89.9k Algerian Darija Conversation Examples for Instruction Tuning

Around 89.9k conversation examples for instruction tuning models in Algerian Darija, a dialect characterized by code-switching between Arabic, French, and local expressions. The dataset was created by the awras-ai project and is hosted on Hugging Face. It was last updated on 2026-06-18.

Use Cases

• Instruction-tune language models for Algerian Darija based on the conversational examples.
• Study linguistic code-switching patterns between Arabic and French based on the described dialect characteristics.
• Develop conversational AI assistants for Algerian users based on the natural response examples.
• Benchmark model performance on low-resource dialectal Arabic based on the provided train/eval split.

Strengths

• Contains around 89.9k conversation examples, providing a substantial corpus for training.
• Includes a predefined split of 80.9k examples for training and 9k for evaluation.
• Specifically designed to capture the natural code-switching typical of everyday Algerian Darija.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

awras (author) as part of the awras-ai project.

Collection Method

Likely gathered and curated for instruction tuning, but specific collection method is not detailed.

Freshness

Last updated 2026-06-18 19:02:23; freshness should be verified.

Geography

Algeria (inferred from the focus on Algerian Darija).