Skip to Main Content
Table 12 

Results of human evaluation performed via Amazon Mechanical Turk. The percentages represent the portion of sentences for which one system had more preference judgments than the other system. If a sentence had an equal number of judgments for the two systems, it was counted in the final row (“neither preferred”).

% of sentences preferred

ZH→EN
UR→EN
Moses, SSVM reranking 33.4% 28.6% 
QPD, TgtTree + TreeToTree 40.6% 42.8% 
neither preferred 26.0% 28.6% 
% of sentences preferred

ZH→EN
UR→EN
Moses, SSVM reranking 33.4% 28.6% 
QPD, TgtTree + TreeToTree 40.6% 42.8% 
neither preferred 26.0% 28.6% 
Close Modal

or Create an Account

Close Modal
Close Modal