A Digitized Corpus of Veps Folktales: Thematic Classification and Applications for an Endangered Finno-Ugric Language (vepsian language)

Veps, an endangered Finno-Ugric language with approximately 4,000 speakers, lacks comprehensive digital resources, limiting its revitalization and computational study. This paper presents a novel corpus of 55 Veps folktales (sourced from the Veps Corpus, ID 767–821, http://vepsian.krc.karelia.ru/text/#subcorpus_Tales), categorized into six thematic clusters to preserve cultural heritage and enable linguistic and NLP applications. As a native Veps speaker and language educator, the author ensured community-aligned development. We propose a baseline algorithm for automated thematic classification, achieving 82% alignment with manual annotations. The corpus, annotated with motifs and translations, supports language revitalization, educational tools, and NLP tasks such as narrative parsing and cross-lingual transfer. It will be shared via LRE Map under a Creative Commons license, contributing to the preservation of Veps and low-resource language research.