Hieu Hoang, University of Edinburgh, UK
This seminar has 2 parts:
1: Improvements to the Statistical Machine Translation toolkit called Moses which is now 7 years old. This toolkit dominates the open source SMT space, is the de-facto standard for phrase-based model and widely adopted by academia. 2: Mixed-syntax model to improve machine translation. Combining the specificity of syntactic models and the generality of the hierarchical phrase-based model.