Daejeon, July 27–31
The concept of the literati (bunjin) is central to the intellectual society of East Asia. The intellectual landscape of early modern East Asia is typically characterized by centralized networks formed through rigid civil service examination systems or strict master-disciple hierarchies, as observed in Korea and China. However, in early modern Japan, which lacked this civil service examination system, diverse classes—including Confucian scholars, doctors, and merchants—became the bearers of this culture. As Nakamura (1958) points out, being a bunjin in 18th-century Japan was not a profession or a class, but a "way of life" based on one's attitude toward literature and art. Reflecting this ethos, the literati community of 18th-century Osaka developed a distinctive social structure independent of such institutions. This community was characterized by a flat, decentralized, and spontaneous ecosystem driven by shared aesthetic pursuits, such as painting, book collecting, and sencha tea ceremonies, which were practiced at physical gatherings known as gashū. At the center of this vibrant community stood the merchant-scholar Kimura Kenkadō (1736–1802). Although Kenkadō held no official political power, he functioned as a vital human node, attracting literati and artists from across the Japanese archipelago.
In the broader context of the "network turn" in the Digital Humanities, mapping decentralized intellectual communities has become an established methodology, exemplified by Stanford University's "Mapping the Republic of Letters." In East Asia, large-scale biographical databases like the China Biographical Database (CBDB) and the Japan Biographical Database (JBDB) are advancing macro-level prosopographical research. Concurrently, textual structuring is progressing in Japan, with notable projects such as the "SAT Daizōkyō Text Database" and the recently released "TEI Classical Text Viewer" (2024), which supports East Asian specificities. However, despite Kenkadō's historical importance, current digital methodologies for analyzing this community remain underdeveloped. While macro-scale databases excel at aggregating flat "who knows whom" edges, they are limited in their ability to capture the situated, micro-dynamics of gashū. Existing resources rely primarily on simple name indexes or flat databases that record "who visited whom." When these datasets are visualized using Social Network Analysis (SNA), they inevitably produce an artifactual "star-shaped" ego-network where all edges connect solely to the author (Kenkadō). Yet, this topology is misleading. It obscures the autonomous nature of the community by masking the rich lateral ties among guests and failing to capture the geographical dynamics of their movements.
Building upon the author's previous studies on the semantic structuring and event-based data modeling of 18th-century Japanese literati poetry societies (Jeong & Kim, 2022; Jeong, 2024), this study aims to deconstruct this author-centric bias. By applying a containment-based Nexus Point semantic TEI modeling approach to Kenkadō Nikki (Kenkadō's diary), this study reconstructs the invisible lateral connections and spatial flows within physical gatherings. This paradigm shift from a "star" to a "mesh" topology proposes a preliminary structural model for the "Japanese-style Literati Network."
To achieve the transition from a "star" to a "mesh" topology, this study shifts the analytical focus from simple "person-to-person visits" to the shared "event" (gashū) itself. To realize this, this study designs and implements a semantic TEI schema that explicitly integrates interpersonal interactions with geospatial data.
While acknowledging the pioneering contribution of Bingenheimer et al. (2011) in establishing the "Nexus-point" model via pointer-based <link> elements, this study proposes a distinct containment-based approach using the TEI <seg> element to suit the situated nature of diaries. By encapsulating a specific interaction (e.g., a morning outing or an evening banquet) within a <seg type="event"> tag, this study captures the immediate context preserved within that text segment. The core value of this semantic model lies in treating each physical event as a structural unit that connects all co-participants. This structure allows subsequent SPARQL queries to algorithmically infer horizontal ties among all co-participants of the event, thereby addressing the flat "star-shaped" bias.
A critical challenge in diary analysis is the frequent omission of the author ("I") from the narrative. To prevent disconnected edges, this study implements an interpretative intervention: explicitly encoding the author using a specialized tag (type="implied"). The default interpretative parameters are defined using <editorialDecl> within the <teiHeader>. This declares that the implied presence of the author defaults to a high level of certainty (@cert="high") and assigns the responsibility to the domain expert (@resp="#editor1"). Specific attributes like @cert="medium" are applied only locally as exceptions for historically ambiguous situations. This explicit encoding strategy ensures the formation of complete "cliques" (triangles) within the graph, providing a consistent structural basis for subsequent SNA.
To ensure data consistency and interoperability, this study adopts a standoff markup strategy. Instead of embedding metadata directly into the text, this study maintains external registries for entities. Person names are linked to VIAF (Virtual International Authority File) to resolve historical variants, while place names are assigned normalized coordinates (WGS84) in an external gazetteer to handle spatial ambiguity. This separation facilitates seamless integration into Linked Open Data (LOD). By keeping the source text structurally clean while directly pointing to global URIs, this approach is compatible with semantic web standards. Consequently, this explicitly enables future SPARQL endpoints to cross-reference the event data of this study with external macro-level databases, supporting scalability and reusability.
Aligning with the sub-theme "Interpretation with Small Data," this study conducts a micro-analysis on a targeted segment (June 1779) of the diary to validate the workflow. This study first constructs Standoff registries linking the entities to global LOD hubs.
Figure 1: TEI markup of the Kenkadō Nikki (June 22–29, 1779)①
Figure 2: TEI markup of the Kenkadō Nikki (June 22–29, 1779)②
Figure 3 TEI markup of the Kenkadō Nikki (June 22–29, 1779)③
Figure 4: SPARQL Query based on the TEI markup of the Kenkadō Nikki (June 22–29, 1779)
The pilot implementation indicates that this event-based model effectively transforms the data structure. By explicitly encoding the implied author and encapsulating co-participants within a <seg>, the preliminary visualizations suggest a shift from a radial star structure to a mesh-like topology. This structural change enables the algorithmic detection of lateral ties between guests (e.g., Ito and other concurrent visitors) independent of the author's explicit mention. Furthermore, the integration of normalized spatial data (e.g., linking "Bishū" to specific coordinates) demonstrates the feasibility of visualizing the geographical range of the network.
Importantly, the semantic attributes encoded in the TEI directly translate into visual parameters in the final Social Network Analysis (SNA) stage. Rather than merely showing binary (presence/absence) connections, the event subtype (e.g., distinguishing a simple visit from a prolonged visit_meal or banquet) is mapped to edge weights in the network graph. This makes it possible to quantitatively measure and visualize the "depth" of interactions. Meanwhile, the @resp attribute ensures that the provenance of all inferred nodes remains transparent.
While this pilot study focuses on a short segment of the diary, a longer-term objective is to construct a "knowledge graph" through a cross-resource analysis extending to additional sources, such as the Zaisinkiji (poetry society records) and Kenkadō's extensive correspondence. By designing a domain-specific ontology, future research will extract and cross-reference information—such as "persons, interaction acts, hobbies, time, and place"—across these heterogeneous sources. Through the conversion workflow from event-based TEI to RDF graph data, validated via SNA, this approach moves beyond a Kenkadō-centric perspective to objectively identify latent hubs within the community and quantitatively measure the depth of interactions. Ensuring LOD compatibility through standoff markup demonstrates the model's scalability for interlinking with external macro-level databases. This methodological approach is expected to help articulate the "Japanese-style Literati Network Model." At the upcoming conference, this presentation aims to demonstrate the SPARQL queries and SNA visualizations derived from this markup, and to discuss the rationale behind the encoding decisions and how this digital approach expands upon existing non-digital historical research.Furthermore, this design demonstrates that a focused, document-centric TEI workflow conducted by a single researcher can generate interoperable data from the outset—serving as a reproducible model for other researchers working with diary and correspondence materials on a comparable scale.