Output details
15 - General Engineering
University of Cambridge
Hierarchical Phrase-based Translation Representations
Joint work with Google Research (New York), supported by a Google Faculty Research Award. EMNLP is a top international conference, with a 24% acceptance rate in 2011. One reviewer noted 'This is a superb paper that takes an automata-theoretic view of the decoding problem in machine translation' and another that 'Rarely does one see a paper heavy with formal language theory, definitions, and algorithms that ends with a strong experimental section, including concrete improvements in (translation) score.' Algorithms described are now implemented in the open source OpenFST Toolkit distributed by Google. A journal article is under review at Computational Linguistics.