@article{oai:ipsj.ixsq.nii.ac.jp:00010263,
 author = {土居, 誉生 and 隅田, 英一郎 and Takao, Doi and Eiichiro, Sumita},
 issue = {6},
 journal = {情報処理学会論文誌},
 month = {Jun},
 note = {コーパスベース翻訳は高い翻訳品質を実現しうる有望な技術である．しかし，入力文と関連のない語が訳文に湧き出すなど，特徴的な翻訳誤りも観察される．本稿では，この湧き出し誤りへの対処として，後編集により自動修正するアプローチを提案する．提案手法は，単語翻訳モデルを利用して誤り語候補を検出し，修正処理を起動する．誤り語候補は，対訳コーパスから得られた用例を利用した制約の検証を経て，訳文から削除される．日英および中英翻訳を対象とした複数のシステムを使った実験で，誤り語自動削除による翻訳精度向上効果が確認された．, Although corpus-based machine translation is a promising technology, one error that is typically observed is out-of-the-blue words, i.e., words that appear in the translation that are unrelated to the input information. Our approach to correct the error is an automatic postedit of the translations. The proposed method finds error clues based on a lexicon model and triggers the correction process. The error words derived from the clues are deleted from the translation after a constraint check using translation examples. We conducted an experiment using several translation systems on Japanese-to-English and Chinese-to-English translations, whose results showed improvement on translation accuracy by automatically deleting unrelated words.},
 pages = {1742--1752},
 title = {単語翻訳モデルを用いた翻訳後編集による湧き出し語対策},
 volume = {47},
 year = {2006}
}