Maloba

Text corpus · Reference version

Dataset-Lingala-Alpaca-LLM

5,700 training examples and 964 test examples, in instruction–response format, across five registers.

View on Hugging Face (opens in a new window)All corpora

How it was built

  • 5,700 training examples and 964 test examples, in instruction–response format, across five registers: translation, formal, urban, educational, summary.
  • Source texts reused under several instructions, to teach the model to change register.

The corpus, as charts

Corpus sizes comparedThe project's four reference corpora, on the same scale. The speech corpus dominates in number of items; translation, the smallest, is the one that moved its model the most.
05 06710 13415 20120 268Speech — segments18 767Text — annotated examples4 818Documents — examples467Translation — human pairs399

Source : Livrable 4, chiffres arrêtés au 2026-09-30

Show the chart data
CorpusItems
Speech — segments18 767
Text — annotated examples4 818
Documents — examples467
Translation — human pairs399

Limits

  • Human validation: 58 % of test examples judged correct by majority, 98 % correct or partial; agreement between validators is weak.

Citation

Congo Digital Services (2026). Dataset-Lingala-Alpaca-LLM. Hugging Face.