次へ: Examination of Word-based Statistical
上へ: Discussion
戻る: Comparison of Automatic Evaluation
Although the B-rank compound/complex sentences were good translations, the simple sentences of the B-rank were worse. We thought that this was because there were fewer English-Japanese word pairs in the simple sentences for making the translation patterns. Thus, there were fewer variables and many characters were left in the compound/complex translation patterns. Therefore, when the input sentence matched an English translation pattern, we obtained a high- quality translation even though the number of matched sentences decreased.
平成25年9月17日