 |
ML Newsnew | past | best | rss | submit | about |
| Unbabel's Participation in the WMT19 Translation Quality Estimation Shared Task (arxiv.org) | |
| 2 points by sel1 on Jul 26, 2019 | hide | past | pdf | discuss on HN | |
|
In plain words: They built a system that guesses how good a machine translation is, word by word or for whole sentences and documents, by combining several prediction models with knowledge borrowed from pre-trained language models. Their entries won every track and language pair by a wide margin. AbstractWe present the contribution of the Unbabel team to the WMT 2019 Shared Task on Quality Estimation. We participated on the word, sentence, and document-level tracks, encompassing 3 language pairs: English-German, English-Russian, and English-French. Our submissions build upon the recent OpenKiwi framework: we combine linear, neural, and predictor-estimator systems with new transfer learning approaches using BERT and XLM pre-trained models. We compare systems individually and propose new ensemble techniques for word and sentence-level predictions. We also propose a simple technique for converting word labels into document-level predictions. Overall, our submitted systems achieve the best results on all tracks and language pairs by a considerable margin. | |
| add comment on HN |