Norn and Nynorn Lexical-Text Corpora for Computational Linguistics

Norn and Nynorn Lexical-Text Corpora Overview This repository contains curated corpora for the extinct North Germanic language Norn and its modern reconstruction Nynorn. The datasets were assembled to support research in: Computational linguistics Natural language processing (NLP) Low-resource language processing Historical linguistics Multilingual word embeddings Language documentation The corpora were created as part of an undergraduate thesis at Waseda University investigating multilingual unsupervised word embedding alignment for extremely low-resource languages. Contents norn_corpus.txt Contains approximately 15,000 tokens consisting of: Lexical headwords extracted from Jakob Jakobsen's Etymological Dictionary of the Norn Language in Shetland Usage examples Quoted Norn and Shetlandic phrases The Ballad of Hildina Shetland Norn and Orkney Norn versions of the Lord's Prayer nynorn_corpus.txt Contains approximately 4,000 tokens consisting of: Nynorn lexical entries Greeting texts Papey Holm poem Purpose These corpora were designed for computational experiments involving: fastText training Multilingual embedding alignment Lexical comparison Historical language reconstruction Although relatively small, they represent one of the largest publicly reusable computational resources currently available for Norn and Nynorn. Sources Primary historical source: Jakob Jakobsen.*Etymological Dictionary of the Norn Language in Shetland.*Volumes I–II (1908–1921). Additional sources include: George Low's transcription of the Hildina Ballad (1774) Norn Lord's Prayer variants Publicly available Nynorn materials Limitations These datasets are hybrid lexical-text corpora. They contain both dictionary-derived lexical material and continuous texts. Because Norn survives only in fragmentary documentation, the corpora are considerably smaller than modern NLP training corpora and should therefore primarily be viewed as resources for low-resource language research. Citation If you use these corpora, please cite the Zenodo DOI associated with this release. License CC BY 4.0