次へ: Proposed Method
上へ: Pattern-Based English-Japanese Machine Translation
戻る: English-Japanese Pattern-Based Machine Translation
GIZA++[3] gets the source language and the target language word pairs by using the maximum likelihood correspondence from the source sentence and target sentence pairs. It also assigns a translation probability. GIZA++ is implemented with IBM model 1-5[2]. In this experiment, we used GIZA++ to obtain the English-Japanese word pairs and the Japanese-English word pairs.
0
表:
Example of English-Japanese Word Pairs with GIZA++
|
|
表:
Example of Japanese-English Word Pairs with GIZA++
|
|
平成25年9月17日
![\scalebox{1.0}[1.0]{
\begin{tabular}{\vert l\vert l\vert l\vert}
\hline
fire & 火 & 0.37\\ \hline
kitchen & 台所 & 0.49\\ \hline
\end{tabular}}](img11.png){kind=small}
![\scalebox{1.0}[1.0]{
\begin{tabular}{\vert l\vert l\vert l\vert}
\hline
火 & fire & 0.22\\ \hline
台所 & kitchen & 0.71\\ \hline
\end{tabular}}](img12.png){kind=small}