Contributing
RAN isn't a registry, and this site isn't its official home. There's deliberately no single authoritative database of RAN scores. The whole point of the method is that it's reproducible: anyone can compute a score for any language from public sources and arrive at the same answer.
The languages collected here are a sample. They exist to show how scores are computed, what they look like across the resource spectrum, and how to read them.
That said, if you'd like a language included in this sample (or a score here corrected) send the following information to me at jared.coleman@lmu.edu:
- The language: its name and ISO 639-3 code.
- The component you're adding or correcting: speakers (S), monolingual (M), or a bilingual pair (Li-Bi).
- A raw count and a source. a number plus where it comes from (Ethnologue for speakers, OSCAR for monolingual, OPUS for a parallel corpus, or another citable source). RAN scores are only as trustworthy as the sources behind them, so a score without a source can't be added.
- A note, optionally: anything about dialects, macrolanguage scope, or why a published figure is wrong.
I'll check the number against the method, recompute the affected digit, and update the language's page. Because every page lists the raw counts behind its digits, corrections are easy to verify after the fact.
Citing RAN
RAN is described in a short paper that will appear at AmericasNLP 2026 from the Kubishi Research Group at Loyola Marymount University (PI: Jared Coleman). If you use the notation or this dataset, please cite that paper.