Model-Based Quality Assessment for Massively Multilingual Parallel Data
Abdelaziz M.A. Ibrahim, Zihao Li, Jörg Tiedemann, Shaoxiong Ji · 2026
Read Model-Based Quality Assessment for Massively Multilingual Parallel Data by Abdelaziz M.A. Ibrahim, Zihao Li, Jörg Tiedemann, Shaoxiong Ji (2026) 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-2606-00285.txt.
Description
Ibrahim and co-authors split the health check of huge dual-tongue data sets in two. Such sets pair lines in two tongues for training use. Faults come in two kinds: pairs that do not match, and pairs with poor craft. One check tool for both can blur the two.
Their check splits pair match from craft grade with no gold text at hand. Four code models are benched on match hunts in two sets. The hunt spans thousands of tongue paths in their stock. Nine craft tools face pro-made lines from the FLORES-200 set. No model wins across all paths, they find. Blends of tools wash out strong marks. Logged tongue reach tracks high craft marks well. They urge path-wise routes in place of one world tool.
Catalog record
| CREATOR | Abdelaziz M.A. Ibrahim, Zihao Li, Jörg Tiedemann, Shaoxiong Ji |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 61 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.00285