# Unbabel's Participation in the WMT19 Translation Quality Estimation Shared Task

## TL;DR

This paper details Unbabel's participation in the WMT19 Quality Estimation shared task, where they used advanced transfer learning and ensemble methods to achieve top results across multiple language pairs and levels.

## Contribution

The paper introduces novel ensemble techniques and transfer learning approaches using BERT and XLM for quality estimation in machine translation.

## Key findings

- Achieved the best results on all tracks and language pairs
- Demonstrated effectiveness of transfer learning with BERT and XLM
- Proposed a simple method for document-level prediction from word labels

## Abstract

We 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.