Modality Matching Matters: Calibrating Language Distances for Cross-Lingual Transfer in URIEL+

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Ng, York Hay, Khan, Aditya, Lu, Xiang, Salloum, Matteo, Zhou, Michael, Hoang, Phuong H., Doğruöz, A. Seza, Lee, En-Shiun Annie
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910064236822528
author Ng, York Hay
Khan, Aditya
Lu, Xiang
Salloum, Matteo
Zhou, Michael
Hoang, Phuong H.
Doğruöz, A. Seza
Lee, En-Shiun Annie
author_facet Ng, York Hay
Khan, Aditya
Lu, Xiang
Salloum, Matteo
Zhou, Michael
Hoang, Phuong H.
Doğruöz, A. Seza
Lee, En-Shiun Annie
contents Existing linguistic knowledge bases such as URIEL+ provide valuable geographic, genetic and typological distances for cross-lingual transfer but suffer from two key limitations. First, their one-size-fits-all vector representations are ill-suited to the diverse structures of linguistic data. Second, they lack a principled method for aggregating these signals into a single, comprehensive score. In this paper, we address these gaps by introducing a framework for type-matched language distances. We propose novel, structure-aware representations for each distance type: speaker-weighted distributions for geography, hyperbolic embeddings for genealogy, and a latent variables model for typology. We unify these signals into a robust, task-agnostic composite distance. Across multiple zero-shot transfer benchmarks, we demonstrate that our representations significantly improve transfer performance when the distance type is relevant to the task, while our composite distance yields gains in most tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2510_19217
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Modality Matching Matters: Calibrating Language Distances for Cross-Lingual Transfer in URIEL+
Ng, York Hay
Khan, Aditya
Lu, Xiang
Salloum, Matteo
Zhou, Michael
Hoang, Phuong H.
Doğruöz, A. Seza
Lee, En-Shiun Annie
Computation and Language
Existing linguistic knowledge bases such as URIEL+ provide valuable geographic, genetic and typological distances for cross-lingual transfer but suffer from two key limitations. First, their one-size-fits-all vector representations are ill-suited to the diverse structures of linguistic data. Second, they lack a principled method for aggregating these signals into a single, comprehensive score. In this paper, we address these gaps by introducing a framework for type-matched language distances. We propose novel, structure-aware representations for each distance type: speaker-weighted distributions for geography, hyperbolic embeddings for genealogy, and a latent variables model for typology. We unify these signals into a robust, task-agnostic composite distance. Across multiple zero-shot transfer benchmarks, we demonstrate that our representations significantly improve transfer performance when the distance type is relevant to the task, while our composite distance yields gains in most tasks.
title Modality Matching Matters: Calibrating Language Distances for Cross-Lingual Transfer in URIEL+
topic Computation and Language
url https://arxiv.org/abs/2510.19217