Builds a WMT18-like corpus for word-level QE with annotations in the source and target words. - View it on GitHub