Giving the Old a Fresh Spin: Quality Estimation-Assisted Constrained Decoding for Automatic Post-Editing
Sourabh Deoghare, Diptesh Kanojia and Pushpak Bhattacharyya · 2025
Read Giving the Old a Fresh Spin: Quality Estimation-Assisted Constrained Decoding for Automatic Post-Editing by Sourabh Deoghare, Diptesh Kanojia and Pushpak Bhattacharyya (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-2501-17265.txt.
Description
Sourabh Deoghare, Diptesh Kanojia, and Pushpak Bhattacharyya curb a bad habit of translation post-editors: fixing what was never broken. Their 2025 paper works on automatic post-editing, where a model polishes machine translation output. Over-correction violates the minimal-edit rule and can harm good text.
Their method feeds word-level quality estimates into the decoder, steering it away from needless changes. The trick is architecture-agnostic, bolting onto any post-editing system. On English-German, English-Hindi, and English-Marathi pairs, error rates improved by 0.65 to 1.86 TER points over matched baselines. The gains show quality estimation and post-editing are natural partners, each supplying what the other lacks. The authors frame restraint as a learnable signal, not a fixed rule.
Catalog record
| CREATOR | Sourabh Deoghare, Diptesh Kanojia and Pushpak Bhattacharyya |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 30 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-17265