Maloba

Translation corpus · Reference version

maloba-llm-alpaca-harmonise

399 French → Lingala pairs translated by Lingala speakers, harmonised spelling.

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

How it was built

  • 399 French → Lingala pairs translated by Lingala speakers (trd configuration): 288 for training, 32 for validation, 79 for test.
  • Transcriptions harmonised by the same program as the speech corpus; the original version is kept in maloba-llm-alpaca.

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

  • A 79-sentence test set: differences are measured by paired bootstrap and confirmed on FLORES+ (1,012 sentences).
  • Adding 4,700 machine-produced pairs changed nothing: only human translations move the model forward.

Citation

Congo Digital Services (2026). maloba-llm-alpaca-harmonise. Hugging Face. NOODL-1.0 licence.