The FBK Participation in the WMT 2016 Automatic Post-editing Shared Task
By Rajen Chatterjee, J. G. D. Souza, Matteo Negri, Marco Turchi. 2016
10 citations recorded by the source corpus.
In this paper, we present a novel approach to combine the two variants of phrase-based APE ( monolingual and context-aware ) by a factored machine translation model that is able to leverage benefits from both. Our factored APE models include part-of-speech-tag and class-based neural language models (LM) along with statistical word-based LM to improve the fluency of the post-edits. These models are built upon a data augmentation technique which helps to mitigate the problem of over-correction in phrase-based APE systems. Our primary APE system further incorporates a quality estimation (QE) model, which aims to select the best translation between the MT output and the automatic post-edit. According to the shared task results, our primary and contrastive (which does not include the QE module) submissions have similar performance and achieved significant improvement of 3.31% TER and 4.25% BLEU (relative) over the baseline MT system on the English-German evaluation set.
Subfields: Artificial Intelligence, 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 Artificial Intelligence
- Ondrej Bojar
- Philipp Koehn
- Lucia Specia
- Graham Neubig
- Matt Post
- Barry Haddow
- Felix Stahlberg
- Rico Sennrich
In Natural Language Processing
- Ondrej Bojar
- Philipp Koehn
- Lucia Specia
- Barry Haddow
- Matt Post
- Christof Monz
- Yvette Graham
- Graham Neubig
Citations in the corpus
- Findings of the 2016 Conference on Machine Translation (2016, background): “, 2016) DCU Dublin City University, Ireland FBK Fondazione Bruno Kessler, Italy (Chatterjee et al., 2016) JUSAAR Jadavpur University, India & Saarland University, Germany USAAR Saarland University, Germany (Pal et al.”
- Findings of the 2017 Conference on Machine Translation (WMT17) (2017, methodology): “Building on previous participations based on the phrase-based paradigm (Chatterjee et al., 2015a, 2016), and similar to (Libovický et al., 2016), such systems jointly learn from source and target information in order to increase robustness and precision of the automatic corrections.”