Kalamang language

Kalamang, sometimes also called Karas, is a divergent Trans–New Guinea language spoken on the biggest of the Karas Islands off the Bomberai Peninsula, that is part of the West Bomberai family.

It was chosen because of its negligible presence in the Internet and because field research materials were collected by Eline Visser, who published "A grammar of Kalamang" as her PhD thesis.

Although Kalamang is primarily oral language, it can be written in the Indonesian alphabet.

Researchers used all existing materials (grammar book, short dictionary, and small set of Kalamang-English sentences) to test how large language models (LLM) can learn a language from a single source, and tested the quality of translations.

[5] In 2024, researchers from Google showed that their latest LLM, Gemini 1.5, can translate English to Kalamang with similar quality to a human who learned from the same resources.