Supervisor

Suitable for

Abstract

Implicit CFG projections for Synchronous Tree-Substitution Grammars (may not be possible) Synchronous Tree Substitution Grammars (STSG) are one of the most promising approaches for next generation machine translation systems. However their complexity means they are very slow to produce translations and require massive amounts of memory. In this project we'll investigate whether we can extend mono-lingual projections for TSGs to bi-lingual STSGs. If we achieve this we may have a much more compact and efficient way of representing grammar based translation models.