Total: 1
We participated in the OLYMPICS task in IWSLT 2012 and submitted two formal runs using a forest-to-string translation system. Our primary run achieved better translation quality than our contrastive run, but worse than a phrase-based and a hierarchical system using Moses.