Ensembling Factored Neural Machine Translation Models for Automatic Post-Editing and Quality Estimation

Proceedings of the Conference on Machine Translation (WMT), Volume 2: Shared Task Papers, pages 647–654 Copenhagen, Denmark, September 711, 2017. c⃝2017 Association for Computational Linguistics Ensembling Factored Neural Machine Translation Models for Automatic Post-Editing and Quality Estimation Chris Hokamp ADAPT Centre Dublin City University chokamp@computing.dcu.ie Abstract This work presents a novel approach to Automatic Post-Editing (APE) and Word- Level Quality Estimation (QE) using en- sembles of specialized Neural Machine Translation (NMT) systems. Word-level features that have proven effective for QE are included as input factors, expanding the representation of the original source and the machine translation hypothesis, which are used to generate an automati- cally post-edited hypothesis. We train a suite of NMT models that use different input representations, but share the same output space. These models are then en- sembled together, and tuned for both the APE and the QE task. We thus attempt to connect the state-of-the-art approaches to APE and QE within a single framework. Our models achieve state-of-the-art results in both tasks, with the only difference in the tuning step which learns weights for each component of the ensemble. 1 Introduction Translation destined for human consumption of- ten must pass through multiple editing stages. In one common scenario, human translators correct machine translation (MT) output, correcting errors and omissions until a perfect translation has been produced. Several studies has shown that this pro- cess, referred to as "post-editing", is faster than translation from scratch (Specia, 2011), or inter- active machine translation (Green et al., 2013). A relatively recent line of research has tried to build models which correct errors in MT auto- matically (Simard et al., 2007; Bojar et al., 2015; Junczys-Dowmunt and Grundkiewicz, 2016). Au- tomatic Post-Editing (APE) typically views the system that produced the original translation as a black box, which cannot be modified or inspected. An APE system has access to the same data that a human translator would see: a source sentence and a translation hypothesis. The job of the sys- tem is to output a corrected hypothesis, attempt- ing to fix errors made by the original translation system. This can be viewed as a sequence-to- sequence task (Sutskever et al., 2014), and is also similar to multi-source machine translation (Zoph and Knight, 2016; Firat et al., 2016). However, APE intuitively tries to make the minimum num- ber of edits required to transform the hypothesis into a satisfactory translation, because we would like our system to mimic human translators in at- tempting to minimize the time spent correcting each MT output. This additional constraint on APE models differentiates the task from multi- source MT. The Word Level QE task is ostensibly a simpler version of APE, where a system must only decide whether or not each word in an MT hypothesis be- longs in the post-edited version – it is not neces- sary to propose a fix for errors. Most recent work has considered word-level QE to be a sequence la- beling task, and employed the standard tools of structured prediction to …

Content truncated to 500 words