Untangling the Influence of Typology, Data and Model Architecture on Ranking Transfer Languages for Cross-Lingual POS Tagging
Enora Rice, Ali Marashian, Hannah Haynie, Katharina von der Wense, Alexis Palmer · 2025
Read Untangling the Influence of Typology, Data and Model Architecture on Ranking Transfer Languages for Cross-Lingual POS Tagging by Enora Rice, Ali Marashian, Hannah Haynie, Katharina von der Wense, Alexis Palmer (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-19979.txt.
Description
Enora Rice and co-authors untangle what makes a good transfer language for grammar tagging in this 2025 paper. Tagging parts of speech in a low-resource tongue can borrow strength from a richer one. Picking the best donor language, though, is still guesswork in practice.
The team trained rankers that score donor languages for a target. Feeds included word overlap, text statistics, family distance, and fine grammar traits. Tests ran across older recurrent models and modern pretrained ones. Word overlap and family distance topped the feature lists everywhere. Mixing grammar traits with dataset facts gave the best rankings overall. Either group alone still ranked respectably, which eases data needs for new languages.
Catalog record
| CREATOR | Enora Rice, Ali Marashian, Hannah Haynie, Katharina von der Wense, Alexis Palmer |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 28 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-19979