行間 Line/Space

2020 Instaration
Collaborator: Tomoya Ishibashi/Ryuji Yoshida

大量の文章を学習したAIが、誰もが知る二文の間に行間を発生させる。
深層学習により、大量の文章をその意味・構造の類似度によってベクトル化した空間を作成。空間内で二文の位置を特定し、それらの文の間に位置する数行の文章を生成させた。
本作では、そのプロセスを可視化するとともに、ベクトルの位置を徐々に移動させたり、異なる文章間を横断させたりすることで、歴史的な言葉を定量的に操作する光景を作り出す。

AI that has learned a large quantity of sentences creates an interlinear space between two sentences that all persons know. Through deep learning, a large amount of sentences were vectorized by their meaning and structural similarity. The position of two sentences in the vector space is specified, and several lines of sentences located between those sentences are generated. In this work, while the process is visualized, by gradually moving the position of the vector or traversing between different sentences, a scene in which historical words are manipulated quantitatively is created.