Giving the Old a Fresh Spin: Quality Estimation-Assisted Constrained Decoding for Automatic Post-Editing

TL;DR

This paper introduces a novel, architecture-agnostic method that uses word-level Quality Estimation to improve automatic post-editing by reducing over-correction, leading to significant translation quality gains across multiple language pairs.

Contribution

It presents a new technique integrating QE into APE decoding to mitigate over-correction, adaptable to any APE system regardless of model or training approach.

Findings

Significant TER improvements on English-German, English-Hindi, and English-Marathi.

Demonstrates the effectiveness of combining QE with APE tasks.

Method is architecture-agnostic and broadly applicable.

Abstract

Automatic Post-Editing (APE) systems often struggle with over-correction, where unnecessary modifications are made to a translation, diverging from the principle of minimal editing. In this paper, we propose a novel technique to mitigate over-correction by incorporating word-level Quality Estimation (QE) information during the decoding process. This method is architecture-agnostic, making it adaptable to any APE system, regardless of the underlying model or training approach. Our experiments on English-German, English-Hindi, and English-Marathi language pairs show the proposed approach yields significant improvements over their corresponding baseline APE systems, with TER gains of $0.65$, $1.86$, and $1.44$ points, respectively. These results underscore the complementary relationship between QE and APE tasks and highlight the effectiveness of integrating QE information to reduce over-correction in APE systems.