AI 中文总结
针对中国传统绘画构图分析难题，引入构图图并开发CompoVista视觉分析系统，通过视觉和上下文查询构建绘画群组，支持多种分析，经评估显示其能助力构图分析，还揭示需求，贡献了结构化表示和分析工作流程。
AI 中文摘要
中国传统绘画（TCP）中的构图承载着空间、叙事和文化美学意义。系统的构图分析对于理解其视觉语言和艺术意义很重要。传统构图分析主要是定性的且由解释驱动，难以跨大型绘画集发现、比较和验证构图模式。为此进行了文献综述和访谈。引入构图图，基于此开发CompoVista视觉分析系统，通过视觉查询和上下文查询构建和修订绘画群组，支持群组级检查、差异比较等。通过用户研究等评估，结果显示其支持构图群组构建等，也揭示了未来需求。该工作为研究TCP构图贡献了特定于构图的结构化表示和集成视觉分析工作流程。
英文摘要
Compositional analysis of Traditional Chinese Paintings (TCPs) reveals how spatial arrangement, narrative structure, and cultural-aesthetic meaning are organized within the pictorial field. Traditional compositional analysis relies primarily on qualitative interpretation, supporting close examination of individual paintings but offering limited capacity to identify, compare, and validate compositional patterns across large-scale collections. To identify the key challenges in analyzing composition across large TCP collections, we collaborated with two art historians and conducted a complementary literature review. Drawing on the resulting insights, we introduce CompoGraph, a structured representation for composition-oriented analysis of TCPs. It represents the composition of a painting across four layers: entities, relations, voids, and context. Based on this representation, we develop CompoVista, a canvas-based visual analytics system for composition-oriented exploration of TCPs. CompoVista allows art historians to construct and refine painting cohorts through interactive compositional queries. It also supports inspecting entity distributions and relations at the cohort level, comparing compositional differences across cohorts, and tracing aggregate patterns back to painting-level evidence. Through two case studies, a user study, and expert interviews, we demonstrate that CompoVista can help art historians discover, compare, and validate compositional patterns across collections of TCPs.
