文脈理解を用いた再生課題の回答正誤による瞳孔変化の分析

摘要

To estimate level of a contextual understanding using oculo-motor indices, two types of sentences such as definition and question statements are shown and they are memorizing and recall tasks. It is obtained that a set of features of eye movements for these statements can predict the recall accuracy including recall rates and correctness of responses in our previous reports, As pupil size was also measured during the tasks as well as eye movements, mean sizes and feature of pupillary changes were analyzed for predicting responses for question statements. While pupil size increased with the number of definition statements, the pupil size decreased with the number of question statements. The phenomenon coincides with a process of short term memorizing and recall task. The mean pupil sizes for 5 and 7 statements conditions were higher than one for 3 statements condition. The information of mean pupil sizes and features of pupillary changes made improve the prediction performance using eye movement features. The prediction performance of correct response for question statements were improved using both eye movement and pupillary features for definition statements in addition to eye movement features for question statements.%眼球の情報によって文脈理解の状態を調べるため，定義文の理解と質問文での強制二肢選択求める文章の記憶再生課題を実施した．この実験においては，定義文での眼球運動の特徴情報から質問文での回答正答率が，質問文での眼球運動の特徴情報から回答正誤がそれぞれ予測できることを示した．本報告では．眼球運動と同時に計測できる瞳孔径の情報を分析し，予測における効果を明らかにした．まず，課題における瞳孔径の変化を調べたところ，定義文を読むごとに瞳孔径が段階的に膨大し．質問文に回答するごとに段階的に縮瞳する現象が見られた．また，提示する定義文の数が5文や7文の条件では，3文の条件よりも瞳孔径が大きかった．定義文の提示時における平均瞳孔径や瞳孔径の変動特徴量を，回答正答率の予測に用いたところ，眼球運動特徴と併用することによって，予測の性能改善が見られた．質問文の回答正誤の予測ではあまり効果は認められなかったが，定義文での眼球運動や瞳孔径の変動特徴量を加えることによって，回答正誤の性能が改善した．