Hybrid Decoding: Decoding with Partial Hypotheses Combination over Multiple SMT Systems

摘要

In this paper, we present hybrid decod-ing — a novel statistical machine transla-tion (SMT) decoding paradigm using mul-tiple SMT systems. In our work, in ad-dition to component SMT systems, sys-tem combination method is also employed in generating partial translation hypothe-ses throughout the decoding process, in which smaller hypotheses generated by each component decoder and hypotheses combination are used in the following de-coding steps to generate larger hypothe-ses. Experimental results on NIST evalu-ation data sets for Chinese-to-English ma-chine translation (MT) task show that our method can not only achieve significant improvements over individual decoders, but also bring substantial gains compared with a state-of-the-art word-level system combination method.