[4I5-OS-17b] OS-17

4:00 PM - 4:15 PM JST(7:00 AM - 7:15 AM UTC)

[4I5-OS-17b-03] Decision Tree Analysis of Artwork ContextualizationA Case Study of the Art Institute of Chicago's Impressionism Collection

〇Shoko Hara¹, Ikki Ohmukai¹ (1. The University of Tokyo)

Keywords:

Contextualization,Decision tree analysis,Metadata analysis,Impressionism,Digital humanities

When museum collections are presented as large-scale image datasets, users often struggle to interpret the underlying organizational structure. This paper addresses this challenge by treating museum metadata as analytical variables and quantitatively identifying the dominant axes of contextualization—the curatorial practice of situating artworks within interpretive frameworks defined by period, region, artist, and technique. We analyze 214 Impressionist works from the Art Institute of Chicago using a two-stage framework: hierarchical clustering with Gower distance and concept-equalized weighting, followed by CART decision tree analysis. Three geographically defined clusters emerge (C1: France, 115 works; C2: United States, 73 works; C3: other European countries, 26 works), with place accounting for 89.1% of variable importance. This finding—robust even after constraining place to one-sixth of total weight—demonstrates that geographic provenance is the dominant structuring principle in this collection, consistent with art-historical accounts of French and American Impressionism as distinct traditions.