04 Oct 2013

Maryam will give a practice talk as well on Friday, October 4th, in ASB 9921. This talk is about Majid and Maryam’s paper in IJCNLP 2013. A short description about the paper is below :

State-of-the-art statistical machine translation systems rely heavily on training data and insufficient training data usually results in poor translation quality. One solution to alleviate this problem is triangulation. Triangulation uses a third language as a pivot through which another source-target translation system can be built. In this paper, we dynamically create multiple such triangulated systems and combine them using a novel approach called ensemble decoding. Experimental results of this approach show significant improvements in the BLEU score over the direct source-target system. Our approach also outperforms a strong linear mixture baseline.