Oral to Web: Digitizing'Zero Resource'Languages of Bangladesh
This study addresses the critical gap in digital resources for approximately 40 Indigenous minority languages of Bangladesh, 14 of which are endangered and predominantly oral with no established writing systems. Through a 90-day field campaign, the project pioneered the simultaneous collection and standardized annotation of 42 language varieties spanning four major language families and unclassified languages. It establishes Bangladesh’s first national multimodal parallel corpus, comprising 85,792 structured text entries and 107 hours of audio recordings transcribed in the International Phonetic Alphabet (IPA), contributed by 77 native speakers. Hosted on the Multilingual Cloud platform, the corpus is publicly accessible and searchable, providing essential infrastructure for systematic digital archiving of zero-resource languages and advancing low-resource natural language processing research.