LIG-CRIStAL System for the WMT17 Automatic Post-Editing Task

TL;DR

This paper introduces two neural post-editing models for the WMT17 task, one with task-specific attention for low-resource scenarios and a chained architecture leveraging source context, showing improved results on en-de and de-en datasets.

Contribution

The paper presents novel neural post-editing models, including a task-specific attention model and a chained architecture, tailored for different data availability scenarios.

Findings

The task-specific attention model performs well in low-resource settings.

The chained architecture slightly improves results with more training data.

Results are demonstrated on en-de and de-en datasets.

Abstract

This paper presents the LIG-CRIStAL submission to the shared Automatic Post- Editing task of WMT 2017. We propose two neural post-editing models: a monosource model with a task-specific attention mechanism, which performs particularly well in a low-resource scenario; and a chained architecture which makes use of the source sentence to provide extra context. This latter architecture manages to slightly improve our results when more training data is available. We present and discuss our results on two datasets (en-de and de-en) that are made available for the task.