Learning from Chunk-based Feedback in Neural Machine Translation

TL;DR

This paper explores how partial, chunk-level user feedback can be effectively used to improve neural machine translation, especially in reducing domain mismatch, showing significant BLEU score improvements.

Contribution

It introduces a simple method to incorporate chunk-based feedback into NMT training and demonstrates its effectiveness over sentence-level feedback.

Findings

Chunk-level feedback outperforms sentence-level feedback by up to 2.61% BLEU.

Using partial feedback reduces domain mismatch in NMT.

Simulation experiments validate the proposed feedback utilization method.

Abstract

We empirically investigate learning from partial feedback in neural machine translation (NMT), when partial feedback is collected by asking users to highlight a correct chunk of a translation. We propose a simple and effective way of utilizing such feedback in NMT training. We demonstrate how the common machine translation problem of domain mismatch between training and deployment can be reduced solely based on chunk-level user feedback. We conduct a series of simulation experiments to test the effectiveness of the proposed method. Our results show that chunk-level feedback outperforms sentence based feedback by up to 2.61% BLEU absolute.