Preference Grammars and Soft Syntactic Constraints for GHKM Syntax-based Statistical Machine Translation
Matthias Huck, Hieu Hoang, Philipp Koehn · 2014
In this work, we investigate the effec-tiveness of two techniques for a feature-based integration of syntactic information into GHKM string-to-tree statistical ma-chine translation (Galley et al., 2004): (1.) Preference grammars on the tar-get language side promote syntactic well-formedness during decoding while also al-lowing for derivations that are not linguis-tically motivated (as in hierarchical trans-lation). (2.) Soft syntactic constraints aug-ment the system with additional source-side syntax features while not modifying the set of string-to-tree translation rules or the baseline feature scores. We conduct experiments with a state-of-the-art setup on an English→German translation task. Our results suggest that preference grammars for GHKM trans-lation are inferior to the plain target-syntactified model, whereas the enhance-ment with soft source syntactic constraints provides consistent gains. By employ-ing soft source syntactic constraints with sparse features, we are able to achieve im-provements of up to 0.7 points BLEU and 1.0 points TER. 1