WMT 2017 Human Evaluations

Pairwise rankings of MT output (2015-2016), and direct assessments (i.e. adequacy and fluency) (2016-2017). In conjunction with the WMT Translation Task Submissions, this can be used for research into MT evaluation. In conjunction with the WMT Translation Task Submissions, this can be used for research into MT evaluation. Numerical data (in csv); 2017 with full output (texts).
Data available here:
http://computing.dcu.ie/~ygraham/newstest2017-system-level-human.tar.gz
http://www.statmt.org/wmt17/results.html

You don’t have the permission to edit this resource.
People who looked at this resource also viewed the following: