Learning Feature Weights using Reward Modeling for Denoising Parallel Corpora

TL;DR

This paper introduces a reward modeling approach to learn feature weights for filtering noisy web-crawled corpora, significantly improving neural machine translation quality across multiple language pairs.

Contribution

It proposes a novel method that learns feature weights optimized for translation performance, surpassing traditional heuristics and single-feature baselines.

Findings

Outperforms strong single feature baselines and hand-designed combinations.

Effective across different noise types and language pairs.

Demonstrates generalization to Maltese-English corpus.

Abstract

Large web-crawled corpora represent an excellent resource for improving the performance of Neural Machine Translation (NMT) systems across several language pairs. However, since these corpora are typically extremely noisy, their use is fairly limited. Current approaches to dealing with this problem mainly focus on filtering using heuristics or single features such as language model scores or bi-lingual similarity. This work presents an alternative approach which learns weights for multiple sentence-level features. These feature weights which are optimized directly for the task of improving translation performance, are used to score and filter sentences in the noisy corpora more effectively. We provide results of applying this technique to building NMT systems using the Paracrawl corpus for Estonian-English and show that it beats strong single feature baselines and hand designed combinations. Additionally, we analyze the sensitivity of this method to different types of noise and explore if the learned weights generalize to other language pairs using the Maltese-English Paracrawl corpus.