Output details
11 - Computer Science and Informatics
University of Oxford
A Gibbs Sampler for Phrasal Synchronous Grammar Induction.
<22>
This paper proposed an approach to learning the synchronous grammars that underpin modern machine translation systems. The Gibbs sampler presented is the first probabilistically correct and scalable inference technique for learning such grammars. The paper includes the largest scale probabilistic synchronous grammar learning experiments published to date. This work directly led to a six week workshop on synchronous grammar induction held at Johns Hopkins University (funded by Google, Microsoft, and DARPA) and underpinned a successful EPSRC First Grant application. ACL is the highest ranked publication venue in Computational Linguistics (Google Scholar Metrics) and ACL2009 had an acceptance rate of 21%.
