1.Introduction

Marriage relationships among villages provide an important lens through which to understand the social structure of traditional Chinese rural communities and the spatial logic of interpersonal communication. Genealogies (zupu族谱) in late imperial China record generations of family members in remarkable details, and the geographical information contained in these texts-particularly the places where men took their wives and where their daughters married outunusually offers rich evidence for reconstructing patterns of spatial interaction. This short paper presents an ongoing project that uses AI-assisted methods to extract and structure marriage-related information from genealogical “shixi” (世系) sections and builds a cross-territory intermarriage network for spatial analysis. By doing so, it re-examines the long-standing scholarly debate over the formation of “marriage circles” in rural China and their relationship to ritual, marketing, and regional systems.

2.History

Since the mid-twentieth century, anthropologists, socialists, geographers and historians have offered contrasting explanations for village intermarriage patterns. Maurice Freedman (Freedman, 1957) and Zhenman Zheng (Zheng, 2001), for example, emphasize the role of lineage organization and ritual spheres, arguing on the basis of southeastern Chinese fieldwork that marriage tended to occur within a ritually defined orbit. In contrast, geographers such as William G. Skinner (Skinner, 1964), and historians including Ishida Hiroshi (Hiroshi, 1980), propose that the “marketing circle”—the spatial field formed by periodic rural markets—served as the primary determinant of marriage distance. Subsequent scholarship has expanded these debates by testing the two models in different regions of China or by proposing hybrid frameworks that incorporate transportation, ecology, local power structures, and economic opportunity. Prasenjit Duara’s research in North China, for example, questions the universality of the marketing-circle model, demonstrating how social and political forces could override market-based spatial logic. (Duara, 2001)

3.Environments and tools

The rapid digitization of genealogical materials in recent years—by libraries, archives, private collectors, and commercial platforms, created new opportunities to revisit these classical debates with more systematic data. The genealogies of Jiangxi province are particularly valuable because many Qing-period texts explicitly record a wife’s original place and a daughter’s marital destination. These entries make it possible to map marriage ties at the scale of the county and to analyze them as spatially embedded social networks. However, the complexity of classical Chinese prose and the irregular nature of genealogical formats make full-scale manual extraction time-consuming and difficult. To address this challenge, this project employs OCR technique by Baimiao app to transform genealogy images to stractured texts, then uses ChatGPT for entity identification, place-name extraction, kinship relation parsing, and structured data generation. Comparing to traditional NER tools, this method allows user-defined rules for text processing, enabling adaptation to domain-specific research requirements. Human-guided verification also remains essential, although AI-assisted preprocessing significantly accelerates the transformation of semi-structured genealogical entries into analyzable datasets. This dataset is not available to share on the Internet because it is part of personal doctoral project and it hasn’t been finished yet.

4.Methodologies

The structured dataset enables the construction of a county-level intermarriage network in which villages function as geographic nodes connected through spousal exchange. By combining Social Network Analysis (SNA) and Geographic Information Systems (GIS), this project measures geodesic marriage distance through ArcGIS geoprocessing tools, identifies network hubs according to betweenness centrality through Gephi calculation, and visualizes spatial flows on digital maps. Village locations were triangulated through local gazetteers, Qing-era maps, and modern mapping services such as Google Maps.

5.Findings

My first analytical step was to revisit the classic conclusion widely cited in earlier anthropological and historical research: that rural marriage circles in traditional China typically fell within a 10 km radius because of limited transportation capacity. The results from this Jiangxi case, however, challenge this assumption. Intermarriage distances in Yongfeng County during the Qing dynasty frequently exceed 20 km, with some routes surpassing 40 km—a finding that indicates a level of mobility that earlier scholarship underestimated. These distances suggest a social world in which mountainous terrain did not necessarily restrict movement to the degree long assumed, and in which local people possessed a more dynamic spatial capacity than modern observers might imagine.

In addition to revising expectations about marriage distance, the network analysis identifies several villages that function as key intermediaries or “bridging” nodes in the marriage structure. Their prominence appears to relate to local social institutions, lineage alliances, transportation routes, and patterns of ecological resource management. These insights invite further exploration into other types of historical materials—especially Qing-period contracts concerning mountainous economic forests and timber rights—which may help determine whether a village’s network prominence is more strongly shaped by marketing activity or by ritual-lineage structures. Such evidence may ultimately clarify how the marketing-circle and ritual-circle models interact in practice and how each contributes to spatial preferences in marriage choice.

6.Value

Methodologically, this project demonstrates how small-scale historical data, when carefully annotated and contextualized, can support interpretive digital humanities research. It also illustrates the value of applying AI techniques to resource-scarce languages and traditional Chinese texts, contributing to broader discussions on data production, algorithmic bias, and the role of humanities-informed small datasets in improving machine learning for specialized domains.