que

que is a privacy-first messaging app working to close the language gap in AI. Large language models perform poorly for speakers of underrepresented languages and varieties, because authentic conversational data in those languages has never been collected at scale. que addresses this at the source: people message privately, and those who choose to can contribute conversation excerpts, with consent and compensation, to build datasets for languages that are largely absent from web-scale training data.

This organization is where we publish dataset samples and schema demonstrations for the research community. We are beginning with Cuban Spanish, a widely spoken but digitally underrepresented variety of a high-resource language, and plan to expand across Latin America and Sub-Saharan Africa.

The packs here are synthetic schema demonstrations for now. Production data is collected only with per-contribution consent and carries a revocable, signed provenance record. We welcome research collaboration, including student and thesis projects.

Web: getque.app · Contact: research@getque.app