Can neural machine translation do simultaneous translation?

TL;DR

This paper explores the use of attention-based neural machine translation for simultaneous translation, introducing a novel decoding algorithm that enables early translation with improved quality by joint segmentation and translation.

Contribution

It presents a new simultaneous greedy decoding algorithm that allows neural machine translation models to start translating before receiving the full source sentence.

Findings

The proposed method enables early translation with maintained quality.

Joint segmentation and translation improve overall translation performance.

First step towards full neural-based simultaneous translation systems.

Abstract

We investigate the potential of attention-based neural machine translation in simultaneous translation. We introduce a novel decoding algorithm, called simultaneous greedy decoding, that allows an existing neural machine translation model to begin translating before a full source sentence is received. This approach is unique from previous works on simultaneous translation in that segmentation and translation are done jointly to maximize the translation quality and that translating each segment is strongly conditioned on all the previous segments. This paper presents a first step toward building a full simultaneous translation system based on neural machine translation.