Improving Neural Machine Translation through Phrase-based Forced Decoding
This work addresses translation quality issues in NMT for language processing applications, but it is incremental as it builds on existing SMT and NMT methods.
The paper tackles the problem of neural machine translation (NMT) sacrificing adequacy for fluency by proposing a method to combine NMT with phrase-based statistical machine translation (SMT) using a soft forced decoding algorithm to rerank NMT outputs, resulting in improved translation quality across four language pairs.
Compared to traditional statistical machine translation (SMT), neural machine translation (NMT) often sacrifices adequacy for the sake of fluency. We propose a method to combine the advantages of traditional SMT and NMT by exploiting an existing phrase-based SMT model to compute the phrase-based decoding cost for an NMT output and then using this cost to rerank the n-best NMT outputs. The main challenge in implementing this approach is that NMT outputs may not be in the search space of the standard phrase-based decoding algorithm, because the search space of phrase-based SMT is limited by the phrase-based translation rule table. We propose a soft forced decoding algorithm, which can always successfully find a decoding path for any NMT output. We show that using the forced decoding cost to rerank the NMT outputs can successfully improve translation quality on four different language pairs.