抄録
本論文では,語彙・構文的言い換えにおいて頻繁に生じる動詞格構造の不整合を自動的に検出する方法を提案する.我々は,コーパスから獲得した大規模な正例に基づいて格構造の適格さを定量化する確率的言語モデルと,人手で収集した小規模な負例に基づいて格構造の不適格さを定量化するモデルを構築し,これら2つを混合し,正例のみに基づく言語モデルと比較して精度の高い誤り検出器を実現した.また,誤り検出に対して貢献度が高い負例を効率良く収集するために能動学習を試行した.
This paper addresses the issue of detecting transfer errors in paraphrasing. Our previous investigation revealed that case assignment of verb tends to be incorrect, irrespective of the types of lexical and structural paraphrasing of Japanese sentences. Motivated by this observation, we propose an empirical method to detect incorrect case assignment. Our error detection model combines two error detection models. They are separately trained on a large collection of positive examples and a small collection of manually labeled negative examples. Experimental results show that our combined model significantly enhances the baseline model which is trained only on positive examples. We also propose a selective sampling scheme to reduce the cost of collecting negative examples, and confirm the effectiveness for the error detection task.
