next up previous
次へ: Compound/Complex Sentences 上へ: Automatic Evaluation of All 戻る: Automatic Evaluation of All

Simple Sentences


We made 10,000 test sentences that combined 379 sentences for the A-rank and 9621 sentences for the baseline(Moses). We called this data "Proposed+Baseline". Next, we evaluated "Proposed+Baseline" and the baseline(Moses).
The results for all test sentences are shown below.

表 XIX: Comparison of All Test Sentences

\scalebox{1.0}[1.0]{
\begin{tabular}{\vert l\vert c\vert c\vert}
\hline
& BLEU...
...1381 & 3.7798\\ \hline
Baseline(Moses) & 0.1375 & 3.7743\\ \hline
\end{tabular}}


In Table XIX, the BLEU score of the "Proposed+Baseline" was higher than the baseline(Moses) by 0.006. This means the "Proposed+Baseline" was more effective than the baseline(Moses).



平成25年9月17日