🇳🇬 Nigerian & Multilingual Language Identifier

All language-identification models by olaverse, in one Space.

Pick a single model to test, or run your text through all of them at once to compare.

Model
Examples

Model notes:

  • lid-neural-5 — Nigerian-focused, 4 languages (Yoruba, Hausa, Igbo, Pidgin)
  • lid-neural-5.1 — Nigerian-focused, same 4 languages, sentence-level tuned on mist-encoder-base-ng (97.6% acc; most residual error involves Pidgin, which shares vocabulary with the others)
  • lid-lite-25 — fastText (character n-gram, CPU-only), 25 languages, two checkpoints (passages / short queries)
  • lid-neural-25.1 / .2 — XLM-R fine-tunes, 25 languages, tuned for passages vs. short queries respectively. Known limitation across both -25 families: Zulu/Xhosa confusion on short text (see model cards).