Paper: Document-Wide Decoding for Phrase-Based Statistical Machine Translation

ACL ID D12-1108
Title Document-Wide Decoding for Phrase-Based Statistical Machine Translation
Venue Conference on Empirical Methods in Natural Language Processing
Session Main Conference
Year 2012
Authors

Independence between sentences is an as- sumption deeply entrenched in the models and algorithms used for statistical machine trans- lation (SMT), particularly in the popular dy- namic programming beam search decoding al- gorithm. This restriction is an obstacle to re- search on more sophisticated discourse-level models for SMT. We propose a stochastic lo- cal search decoding method for phrase-based SMT, which permits free document-wide de- pendencies in the models. We explore the sta- bility and the search parameters of this method and demonstrate that it can be successfully used to optimise a document-level semantic language model. 1 Motivation In the field of translation studies, it is undisputed that discourse-wide context must be considered care- fully for good translation results (Ha...