Ensembling Factored Neural Machine Translation Models for Automatic Post-Editing and Quality Estimation
By Chris Hokamp. 2017
38 citations recorded by the source corpus.
This work presents a novel approach to Automatic Post-Editing (APE) and Word-Level Quality Estimation (QE) using ensembles 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 automatically 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 ensembled 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.
Subfields: Natural Language Processing
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Natural Language Processing
- Philipp Koehn
- Ondrej Bojar
- Lucia Specia
- Graham Neubig
- Luke S. Zettlemoyer
- Barry Haddow
- Rico Sennrich
- Matt Post
Citations in the corpus
- Simple and Effective Retrieve-Edit-Rerank Text Generation (2020, background): “, 2018; Cao and Xiong, 2018) and post-editing for NMT (Hokamp, 2017; Dabre et al., 2017).”
- Findings of the 2017 Conference on Machine Translation (WMT17) (2017, background): “AMU Adam Mickiewicz University, Poland (Junczys-Dowmunt and Grundkiewicz, 2017) CUNI Univerzita Karlova v Praze, Czech Republic (Variš and Bojar, 2017) DCU Dublin City University, Ireland (Hokamp, 2017) FBK Fondazione Bruno Kessler, Italy (Chatterjee et al.”
- Neural Fuzzy Repair: Integrating Fuzzy Matches into Neural Machine Translation (2019, methodology): “nation of source sentence and MT output at the source side, with a specific break token separating the two strings (Hokamp, 2017).”