decoding with syntactic and non-syntactic phrases in a syntax-based machine translation system

A key concern in building syntax-based ma-chine translation systems is how to improve coverage by incorporating more traditional phrase-based SMT phrase pairs that do not correspond to syntactic constituents. At the same time, it is desirable to include as much syntactic information in the system as pos-sible in order to carry out linguistically mo-tivated reordering, for example. We apply an extended and modified version of the ap-proach of Tinsley et al. (2007), extracting syntax-based phrase pairs from a large parallel parsed corpus, combining them with PBSMT phrases, and performing joint decoding in a syntax-based MT framework without loss of translation quality. This effectively addresses the low coverage of purely syntactic MT with-out discarding syntactic information. Further, we show the potential for improved transla-tion results with the inclusion of a syntactic grammar. We also introduce a new syntax-prioritized technique for combining syntactic and non-syntactic phrases that reduces overall phrase table size and decoding time by 61%, with only a minimal drop in automatic trans-lation metric scores.