Thesis
Respeaking for Word-Spotting as a Collaborative Transcription Method for Oral Languages
University of Melbourne 2020
TL;DR
This thesis studies respeaking and word spotting as a collaborative transcription method for oral languages. It focuses on making speech technology useful for language documentation workflows.
Why It Matters
Language documentation often faces scarce transcription resources. Respeaking offers a way to combine human collaboration with speech recognition to support transcription of oral and endangered languages.
Abstract
This thesis explores the use of respeaking and speech recognition to assist in documenting and transcribing oral and endangered languages.
Paper Content
Core Contribution
This thesis studies respeaking as a collaborative transcription method for oral languages. It explores how a speaker can repeat or reformulate audio in ways that make speech recognition and word-spotting tools more useful for transcription.
Research Use
The work is relevant for language documentation, endangered language technology, speech recognition workflows, and collaborative transcription methods where direct automatic recognition is difficult.