Speech embeddings from a fine-tuned language identification model produce language distance measures that align with lexical and geographic distances across 106 languages, with around 50 recordings per language needed for stable results.
Historical Linguistics
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Neighbors and relatives: How do speech embeddings reflect linguistic connections across the world?
Speech embeddings from a fine-tuned language identification model produce language distance measures that align with lexical and geographic distances across 106 languages, with around 50 recordings per language needed for stable results.