This study focuses on Sima Guang's Dule Yuan (Garden of Solitary Joy) in the Northern Song dynasty, a site that functioned both as a personal spiritual emblem for its owner and as a multivalent cultural metaphor in later ages. It presents an interdisciplinary exploration that combines data analysis, cultural interpretation, and AI-based visual experiments. While scholars have long examined the interplay among poetry, painting, and gardens in Song literati culture—revealing the spiritual pursuits, social practices, and historical contexts underlying them—previous studies have often treated such materials as relatively static cultural content, paying insufficient attention to their dynamic transformations across different social networks and historical contexts. This project instead approaches Dule Yuan as a cultural image dynamically constructed through multiple discourses over time—at once an expression of Sima Guang's personal experience and a case study of how Chinese classical cultural symbols acquire new meanings through transmission and re-contextualization.

Drawing upon a corpus of seventy-one poems related to Dule Yuan, the research constructs a three-tiered diachronic narrative: Sima Guang's own compositions, the poetic exchanges of his contemporaries, and the reinterpretations by later literati. At the quantitative level, the study employs word frequency analysis, topic modeling, and semantic network analysis to reveal the semantic migration of the garden's core imagery—for instance, its evolution from notions of "leisure" (xianshi) and "self-contentment" (zide) in Sima Guang's era to those of "moral virtue" (dexing) and "integrity" (fenggu) in later contexts. Subsequently, at the interpretive level, these data-driven patterns are contextualized through a hermeneutic reading of paintings, political history, and Confucian thought, elucidating the cultural drivers behind the garden's transformation into a potent symbol of political and moral significance.

The textual abundance yet lack of visual documentation of Dule Yuan makes it an ideal case study for examining and critiquing how text-to-image models translate complex cultural memory. Accordingly, the research utilizes these models to generate visualizations of Dule Yuan in different historical contexts. This process not only reveals the cultural "misreadings" caused by the limitations of contemporary visual corpora but also thereby confirms the biases of training data and the epistemological limits of the models, highlighting the necessity of humanistic intervention in interpretation and correction. In response, this study proposes a paradigm for AI prompt design rooted in the classical Chinese poetic concepts of xing (form), xiang (image), and yi (idea), aiming to explore an effective visual generation pathway for reproducing cultural memory and a novel model for human-machine collaboration.

In conclusion, this research not only illuminates the dynamic evolution of Dule Yuan as a cultural image but also advances a methodological paradigm that integrates quantitative analysis, cultural interpretation, and AI critique. Beyond offering a new interpretation of Dule Yuan, the project demonstrates how classical Chinese texts can enter into dialogue with global digital humanities, thus contributing to broader international discussions on cultural memory and re-contextualization at the intersection of historical tradition and digital futures.