Neural Machine Translation Decoding with Terminology Constraints

TL;DR

This paper presents a novel constrained decoding method for neural machine translation that incorporates user-provided terminology constraints using finite-state machines and multi-stack decoding, improving translation accuracy and adherence to constraints.

Contribution

The authors introduce a new constrained decoding framework for NMT that supports target-side and input-aligned constraints, enhancing control over translation outputs.

Findings

Effective incorporation of terminology constraints in NMT decoding.

Reduction in misplacement and duplication of constraints.

Demonstrated improvements across multiple translation tasks.

Abstract

Despite the impressive quality improvements yielded by neural machine translation (NMT) systems, controlling their translation output to adhere to user-provided terminology constraints remains an open problem. We describe our approach to constrained neural decoding based on finite-state machines and multi-stack decoding which supports target-side constraints as well as constraints with corresponding aligned input text spans. We demonstrate the performance of our framework on multiple translation tasks and motivate the need for constrained decoding with attentions as a means of reducing misplacement and duplication when translating user constraints.