Confidence-aware semi-supervised siamese graph networks for semantic text similarity in low-resource Urdu
Anil Ahmed, Lei Guo, Degen Huang, Salahuddin Unar, Mobeen Nazar · 2026
Read Confidence-aware semi-supervised siamese graph networks for semantic text similarity in low-resource Urdu by Anil Ahmed, Lei Guo, Degen Huang, Salahuddin Unar, Mobeen Nazar (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/PMC13542180.txt.
Description
Anil Ahmed and four colleagues tackle, in a 2026 study, judging whether two Urdu sentences mean the same thing, a task starved of labelled examples in this low-resource language. Their framework learns partly from unlabelled text it labels itself, but only when it is confident.
Back-translation and word masking manufactured extra training variety. A twin graph network blending two graph methods compared sentence structures, sharpened by an attention encoder that resists the blurring graph networks suffer. Triple filtering by probability, meaning match, and uncertainty kept pseudo-labels clean. Across six benchmark sets the system beat strong rivals, with one dataset showing an eight-point accuracy gain over the best baseline.
Catalog record
| CREATOR | Anil Ahmed, Lei Guo, Degen Huang, Salahuddin Unar, Mobeen Nazar |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 164233 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |