Learning to Prune: Context-Sensitive Pruning for Syntactic MT
Wenduan Xu, Yue Zhang, Philip J. Williams, Philipp Koehn · 2013
We present a context-sensitive chart prun-ing method for CKY-style MT decoding. Source phrases that are unlikely to have aligned target constituents are identified using sequence labellers learned from the parallel corpus, and speed-up is obtained by pruning corresponding chart cells. The proposed method is easy to implement, or-thogonal to cube pruning and additive to its pruning power. On a full-scale English-to-German experiment with a string-to-tree model, we obtain a speed-up of more than 60 % over a strong baseline, with no loss in BLEU. 1