next up previous
次へ: Experiments with our Machine 上へ: Concepts of our Statistical 戻る: Training

Decoding

The decoding process is as follows.

  1. Test Corpus

    We prepare the Japanese test sentences.

  2. Rule-based Machine Translation

    We used a Japanese-English rule-based machine translation. Thus, we obtain "ENGLISH" test sentences.

  3. Statistical Machine Translation System

    Using phrase table in Section 3.1, $ N$ -gram model in Section 3.1, and moses[7], we decode the "ENGLISH" sentences. This involves "ENGLISH"-English translation. In this way, we obtain English sentences.

Fig.2 shows the flow chart of the decoding process.

図 2: Flowchart of Decoding
\fbox{
\includegraphics[width=0.6\columnwidth]{figure/figure2.eps}
}



平成24年1月18日