Most speech technology is built for a handful of widely spoken languages. Thousands of languages, including many Indigenous languages, have little or no recorded speech, few written resources and no working TTS or ASR. I want to help change that, in close partnership with the communities who speak these languages.

Community first

Language data belongs to the people who speak the language. I follow the CARE Principles for Indigenous Data Governance (Collective benefit, Authority to control, Responsibility and Ethics): communities decide what is recorded, how it is used, who has access and what the technology is for. Speakers give informed consent and are fairly compensated.

How I can help

  • Project scoping: what is realistic with the data you have, and what would make the biggest difference for your community or program.
  • Speech data collection: recording scripts, protocols and annotation guidelines designed for small speaker communities, elders and second-language learners.
  • Text-to-speech: voices for language learning apps, dictionaries and teaching materials, including pronunciation lexicons and text normalization built for the language.
  • Speech recognition: transcription support for archival recordings and fieldwork, using multilingual models and low-resource adaptation techniques.
  • Evaluation with speakers: listening tests and feedback sessions with community members, so quality is judged by the people who will use the technology.

Who this is for

Tribal nations and language programs, revitalization organizations, universities and researchers, and companies that want to support more languages responsibly.