Seeding alignment between language technology and indigenous methodologies : the making of Kʷu sqilxʷ an interactive digital edition

This thesis describes the development of an interactive digital edition of Kʷu Sqilxʷ /We are the People: A Trilogy of Okanagan Legends in nsyilxcən. It is the first digital edition to date to use automatic speech-to-text alignment for nsyilxcən. The written portion of this Digital Humanities (DH) project addresses a longstanding schism between western and Indigenous ways of knowing that threaten to further a history of disingenuous settler-colonial research. The overarching research question examines how the development of this digital edition exposes issues of colonial praxis embedded in disciplines of computer science and linguistics. By reviewing the story of the successful implementation of this speech technology in collaboration with syilx language advocates, this thesis hopes to “seed alignment” between speech technologies and Indigenous methodologies. To approach the speech-to-text alignment challenge for a low-resourced language, this project leverages the cross-lingual approach used by the ReadAlongs platform developed by the National Research Council of Canada (NRC). This DH project’s contribution to the ReadAlongs tool was the development of a grapheme-to-phoneme model specifically for nsyilxcən, enabling future projects for nsyilxcən and related Salishan languages. The project braids computational linguistics, digital humanities, and Indigenous methodologies, embracing interdisciplinarity as a decolonising methodology. As a non-Indigenous settler researcher, I emphasize the importance of reflexivity, critical listening positionality, and ethical collaboration with the Indigenous community throughout the research process (Robinson, 2020; Kovach, 2021). By leveraging resources from a language with greater linguistic resources, this project employs what is known as a “zero-shot” technique, an approach which upholds Indigenous Data Sovereignty by avoiding the collection of nsyilxcən language data.