Elpis, an accessible speech-to-text tool

Ben Foley, Alina Rakhi, Nicholas Lambourne, Nicholas Buckeridge, Janet Wiles · 2019

Elpis is a speech-to-text tool which has been designed to give language workers, including linguists and speech scientists, access to cutting-edge automatic speech recognition software, without the specialist training typically required to run these systems. Our presentation would demonstrate local and server-based versions of Elpis using sample data from the Abui (ISO 639: abz) language, about 17,000 speakers in Indonesia. Attendees would gain a sense of the benefits that a first-pass ASR transcription can bring to transcription workflows.

Read the paper · More papers on PaperTik