SESSION ANNULÉE - Using text–to–speech alignment technology to support oral production and comprehension in Mushkegowuk (Swampy Cree)

Session régulière / Regular Session / Sesión regular
11:15 AM, mercredi 22 mai 2019 (40 minutes)

ANNULÉE - Using text–to–speech alignment technology to support oral production and comprehension in Mushkegowuk (Swampy Cree) - ANNULÉE

In this talk, I address issues that arise while developing text-alignment technology to produce read-along as a mediating tool to facilitate Mushkegowuk L2 leaners’ oral comprehension and production. Despite rapid advancements in information and communication technologies for the purpose of language teaching and learning for majority languages, indigenous languages still face challenges incorporating these technologies as a pedagogical tool. To fill this gap, this study aims to produce text-to-speech-alignment from recordings of Mushkegowuk language traditional stories previously collected and transcribed by Dr. Doug Ellis during the period of 1950 to 2010 along the Coast of James Bay. The text-speech-alignment consists of combining two language modalities, written and oral, and determining which segments of audio correspond to words and/or phonemes in the transcription (cf. Littell et al., 2018). However, this process of alignment raises questions about how to establish a relationship between analytical categories that are particular to each modality. That is, the boundaries of a printed word/phrase/sentence/paragraph found in the written text may not correspond to the matching spoken word/utterance/turn in the audio recordings. Thus, this study aims to explore the initial challenges faced during the text-speech-alignment process and to propose a template that captures and preserves the linguistic features of each modality. To achieve this goal, I use multimodal annotation tool ELAN (cf. Crasborn & Sloetjes, 2014) used in language documentation to combine the written text and corresponding sound files. By producing the text-to-speech-alignment the existing audio material will be made available to teachers, learners, and to the community. Using a text-to-speech technology in a community-based research project, I expect to facilitate the use of technologies for teaching, learning and revitalization of the Cree language.

UQAM - École de langues
Carleton University

Mon horaire

Ajouter à Mon Horaire comme Favoris