The first generation of the statistical machine translation was word-based, and its performance was low. More recently, phrase-based statistical machine translation has gotten better results.
In the proposed method, we think that the Japanese-English word dictionary and Japanese-English translation patterns were equivalent to the translation model of SMT. We also think that the word tri-gram was equivalent to the language model of SMT. Consequently, we think that the proposed method was equivalent to a word-based SMT decoder.