The first openly-licensed, code-switch-labeled speech benchmark for Hiligaynon (Ilonggo) — a language spoken by 9M+ Filipinos yet nearly invisible to modern speech technology.

· Code: https://github.com/Jazztinn/tinig-sa-liwanag · License: CC BY 4.0 Real Hiligaynon-English-Tagalog speech, labeled per word, with a scorer that measures what generic models miss: accuracy at the moment the language… See the full description on the dataset page: https://huggingface.co/datasets/LauelKills/sugidanon-hil-codeswitch.