Maloba

Speech, external corpus · Earlier version

Speech Recognition Datasets for Congolese Languages (lingala)

Open-source corpus folded into the Kinshasa Voice-a-thon: 2,940 segments, 5 h 00 min, 25 to 31 speakers.

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

How it was built

  • A corpus from outside the project, taken as it stands: recordings of Congolese languages collected in Kinshasa, Lingala part.
  • 2,940 segments, 5 h 00 min 18 s, 31 speaker identifiers (25 to 31 probable people).

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

  • It does not come from the Maloba collection: it is counted separately and does not enter the project's hours.
  • CC BY 4.0 licence, distinct from the NOODL-1.0 of the Maloba corpora: credit the original authors.

Citation

Speech Recognition Datasets for Congolese Languages, Mendeley Data. CC BY 4.0 licence.