next up previous
次へ: Conclusion 上へ: Statistical Machine Translation with 戻る: Human Evaluation Results

Discussion

With our system, we aim to reduce the number of ungrammatical sentences. Thus, we analyze the outputs according to these factors. However, the patent sentences are too long and too strange sentences compared normal sentences, it was impossible to analyze these results for detail, and could not determine what was wrong. However, by comparing the output of moses and the output of our system, we found that our system produced more grammatically correct sentences.



平成24年1月18日