Ashuza11/drc-bitext-eval

Parallel text collection and translation evaluation (BLEU / chrF++) for low-resource languages of the DRC. # DRC Language Data Collection Worksheets Evidence-audited, copy-ready worksheets for bitexts, bilingual lexicons, and monolingual corpora. Last checked: **2026-07-20**. ## Group 1 — The four national languages | Constitutional name | Data-collection target | ISO 639-3 | Worksheet | |---|---|---:|---| | Lingala | Lingala | `lin` | Lingala | | Ciluba | Ciluba / Tshiluba / Luba-Kasai | `lua` | Ciluba | | Kikongo | **Kikongo ya Leta / DRC Kituba** | `ktu` | Kikongo ya Leta | | Swahili | **Congo Swahili from the DRC only** | `swc` | Congo Swahili | The Constitution names “Kikongo,” but the national lingua franca commonly meant in present-day DRC is Kikongo ya Leta (DRC Kituba, `ktu`). The Kongo macrolanguage code `kon`, Koongo `kng`, and Angolan San Salvador Kongo `kwy` are not interchangeable with `ktu`. The Swahili worksheet excludes generic/standard Swahili (`swa`, `swh`, `sw`). “Kingwana” is retained as a search term for northern/Ituri varieties, not as a synonym for every `swc` resource. ## Group 2 — Eastern DRC regional languages These are languages, not collectively “dialects.” | Language | ISO 639-3 | Core area | Worksheet | |---|---:|---|---| | Mashi / Shi | `shr` | Bukavu hinterland; Walungu, Kabare, Kalehe | Mashi | | Nande / Kinande | `nnb` | Beni and Lubero; also neighboring areas | Nande | | Hunde / Kihunde | `hke` | Masisi, Rutshuru, Walikale and nearby areas | Hunde | | Fuliiru / Kifuliiru | `flr` | Uvira Plain and Uvira Territory | Fuliiru | | Tembo / Kitembo | `tbt` | Kalehe, Masisi and Walikale | Tembo | | Havu / Kihavu | `hav` | Idjwi and Kalehe | Havu | | Nyanga / Kinyanga | `nyj` | Walikale | Nyanga | | Lega-Mwenga | `leg` | Mwenga | Lega | | Lega-Shabunda | `lea` | Shabunda and Pangi | Lega | ## Evidence labels - **VERIFIED DATASET**: files, language identity, access route, and licence checked. - **CONVERTIBLE SOURCE**: real content exists, but alignment, extraction, or permission is needed. - **CATALOG RECORD**: proves a resource exists …