Phrasal Segmentation Models for Statistical Machine Translation
Graeme Blackwood, Adrià de Gispert, Bill Byrne · 2008
Phrasal segmentation models define a mapping from the words of a sentence to sequences of translatable phrases. We discuss the estimation of these models from large quantities of monolingual train-ing text and describe their realization as weighted finite state transducers for incor-poration into phrase-based statistical ma-chine translation systems. Results are re-ported on the NIST Arabic-English trans-lation tasks showing significant comple-mentary gains in BLEU score with large 5-gram and 6-gram language models. 1