Translation model combination based on string-chunk

We present a translation model combination that integrate SCFG with BTG, and introduce a string-chunk model into this combination. It is different from system combination which is postprocessing to n-best result of every translation system, however model combination combine multiple translation model when decoding. Different from joint decoding, we use new single translation model, and in our method multiple machine translation decoders share hypergraph vector with string and chunk information through multi-decoding style. Experimental results on data sets for NIST Chinese-to-English machine translation task show that the translation model combination based on string-chunk method can bring significant improvements to all baseline decoders, and the outputs from translation model combination can be used to further improve the result of system combination.