Software for enriching endangered language-annotated databases with crowd-sourced linguistic and cultural input

University Of Kentucky Research Foundation, Lexington KY

Investigators

Abstract

To understand how humans organize and relay information, linguists need to know what linguistic structures are possible in human language. Yet many languages that can tell us more about potential linguistic structures are no longer being spoken as fluent older speakers are not being replaced by younger speakers, thus making the loss of linguistic and cultural knowledge irreversible. The Documenting Endangered Languages Program supports proposals, including those that create innovative software, to facilitate the efficient, rapid, and accurate digital capture of endangered language data so that this knowledge can be collected, analyzed, and made available for speakers and scientists in perpetuity. With a three-year award Raphael Finkel of the University of Kentucky and Daniel Kaufmann of the Endangered Language Alliance will create software that allows for the rapid refinement of transcription and translation of endangered language data. The team will test their software on data collected on two languages -- Purhepecha, an isolate spoken in Mexico, and Koda, a Munda language spoken in Bangladesh and India. A central product of language documentation projects is the annotated corpus, a collection of language samples with accompanying analytic notes on specific features such as the sounds and word and sentence structure of the language. Kaufman and Finkel will write software that will provide web-based access to the corpus and make it possible for linguists and the speaker-community to search, browse, and edit the corpus. These additional participatory annotations will enrich and refine the corpus in ways and at a rate not previously possible. Thus, in addition to providing valuable data on two under-described languages, the results of this project have the potential of dramatically changing the quality of language documentation products. Data from these projects will be available at the Archive of the Indigenous languages of Latin America and the Endangered Languages Archive and Repository. The resulting software will be distributed through SourceForge.