Audio Classifier for Endangered Language Analysis and Education

Abstract

Around 42% of the world languages are considered endangered due to the decline in the number of speakers. MeTILDA (Melodic Transcription in Language Documentation and Application) is a collaborative platform created for researchers, teachers, and students to interact, teach, and learn endangered languages. It is currently being developed and tested on the Blackfoot language, an endangered language primarily spoken in Northwest Montana, USA and Southern Alberta, Canada. This study extends MeTILDA functionality by incorporating machine learning framework in documenting, analyzing, and educating endangered languages. Specifically, this application focuses on two main components, namely audio classifier and language learning. Here, the audio classifier component allows users to automatically obtain instances of vowels and consonants in Blackfoot audio files. The language learning component enables users to visually study the pitch patterns of these instances and improve their pronunciation by comparing with that of native speakers using a perceptual scale. This application reduces manual efforts and time-intensive tasks in locating important segments of Blackfoot language for research and educational purpose.