Digital resources in the Social Sciences and Humanities OpenEdition Our platforms OpenEdition Books OpenEdition Journals Hypotheses Calenda Libraries OpenEdition Freemium Follow us

Behind the Counter: When Shopkeepers Made (and Lost) the News

In the previous post, I examined the problem of terminology and the different labels that circulated in the Shenbao to designate shopkeepers. That exercise was not an end in itself but a necessary step toward building a corpus. Terminology provided the bridge between abstract categories and the tangible traces left in the pages of the newspaper. By turning words into queries, I was able to extract a coherent set of articles that bring shopkeepers into view. This post takes up the next stage of the process: how choices made at the level of vocabulary translate into the construction of a corpus, and how this corpus becomes the foundation for any further analysis.

The selected vocabulary produced a corpus of 69,851 articles. This scale is striking when set against the total size of the Shenbao archive, which runs into close to three million items over its seventy-seven years of publication. Even at nearly seventy thousand articles, the “shopkeeper corpus” represents only a small fraction of the newspaper’s total output. This underlines both the richness and the limitations of the material: rich, because the corpus is large enough to trace long-term patterns across decades; limited, because vast sections of the newspaper—political commentary, serialized fiction, advertisements—remain outside its scope. In other words, corpus building is as much about what is captured as what is left behind, and our subsequent interpretations must remain attentive to this imbalance.

In this essay, I shall examine successively the corpus of articles found in the Shenbao and in the English-language press.

The visibility of shopkeepers in the Shenbao tells two different stories depending on how we measure it. When we count the raw number of articles mentioning shopkeepers each year, we see a pattern of steady growth culminating in peaks during the 1920s and early 1930s (Fig. 2). But when we calculate what percentage of all articles mentioned shopkeepers, a different narrative emerges: a gradual marginalization as the Shanghai press turned its attention to broader political and international concerns (Fig. 3).

In the early decades of newspaper, from the 1870s through the 1890s, shopkeeper mentions began modestly but climbed steadily upward (Fig. 1). This growth reflected Shanghai’s commercial expansion and the Shenbao‘s increasing coverage of urban economic life. The turn of the century brought more dramatic increases, with noticeable peaks appearing in the late 1900s and early 1910s. These surges coincided with the emergence of Shanghai as China’s main commercial center and the social turbulence that marked the late Qing dynasty’s final years. The 1920s and 1930s witnessed the zenith of shopkeeper coverage in absolute terms. Articles mentioning shopkeepers reached their highest levels during the early 1930s, through the Nanjing decade, when the press devoted sustained attention to commercial affairs. This peak reflected both the maturation of  retail economy in the city and the expanded capacity of the newspaper to cover it. The Shenbao was publishing more articles overall, and shopkeepers featured prominently in this expanding coverage. Then came the precipitous decline of the 1940s. As war engulfed China, mentions of shopkeepers collapsed sharply. Brief spikes punctuated an otherwise bleak landscape of minimal coverage. By the late 1940s, in the final years before 1949, references to shopkeepers had fallen to levels far below those of earlier decades, reflecting the severe constraints on press activity.Yet the temporal curve should not be read too literally as a mirror of shopkeepers’ historical significance. Spikes and declines may reflect not only real changes in commercial life but also shifts in editorial priorities, political censorship, or even the mechanics of keyword searching. Any temporal distribution drawn from corpus queries thus needs to be read as an interplay between historical processes, editorial dynamics, and the structural biases of digital text retrieval.

Figure 1. Distribution of Articles in the Shenbao: Total Number of Articles per Year (1872-1949)

Yet this growth narrative obscures a more complex reality revealed when we examine the share of shopkeepers in total newspaper coverage (Fig. 2). In the 1870s and 1880s, shopkeepers dominated in the Shenbao in a way they would never match again. During these early years, shopkeeper-related articles sometimes exceeded seven percent of all published content. This high proportion reflected the limited thematic scope of early Shanghai journalism. The newspaper had fewer competing subjects, and urban commerce naturally occupied center stage. The proportion remained robust through the 1890s and into the 1900s, fluctuating between five and seven percent. Shopkeepers still commanded significant editorial attention, marking them as central figures in the portrayal of Shanghai life by the newspaper. But by the 1910s and 1920s, a steady erosion had begun. The percentage dropped to around two or three percent, and it continued falling through the 1930s, often dipping below two percent even during years when absolute mentions peaked.

Figure 2. Distribution of Articles in the Shenbao: Proportion of all Articles per Year (1872-1949)

This paradox is striking: during the very decades when the Shenbao published more articles about shopkeepers than ever before, these articles constituted a smaller fraction of the overall output in the newspaper. Other subjects—political upheavals, international affairs, military conflicts, and social movements—claimed an ever-larger share of column inches. Shopkeepers had not disappeared from the news; they had been crowded out by competing narratives. The war years after 1937 brought both absolute and relative decline. Mentions became sporadic and marginal, barely registering as a percentage of total coverage. The commercial world that had once dominated early editions of the Shenbao had receded to the periphery of wartime journalism.

These two divergent trajectories illuminate a fundamental transformation in the local public sphere. The Shenbao itself expanded dramatically over these decades, multiplying its pages and diversifying its coverage. Shanghai grew from a treaty port into a metropolis, and the range of newsworthy events expanded accordingly.[1] Political crises, international tensions, intellectual debates, and eventually war generated stories that competed for space with commercial news. Shopkeepers thus experienced a paradox of visibility. As the retail sector flourished and the press wrote about it more frequently in raw numbers, shopkeepers simultaneously lost their privileged position at the center of urban journalism. They were mentioned more but mattered less, at least in terms of their share of editorial attention. This shift reflects changing conceptions of what constituted important news. In the 1870s and 1880s, when the Shenbao focused heavily on local commercial life, shopkeepers embodied the core focus of the newspaper: documenting the daily rhythms of Shanghai’s economic world. By the 1920s and 1930s, even as the commercial sector reached new heights, the press had redefined its priorities. National politics, international relations, and ideological conflicts overshadowed local economic actors.

When we turn to the English-language press in Shanghai, the trajectory of references to shopkeepers unfolds differently. In terms of sheer volume, the presence of shopkeepers in the English-language press was much more modest: altogether our query produced a corpus of 5,145 articles. The contrast with the Shenbao corpus is immediately evident: the same query structure that yielded nearly seventy thousand results in Chinese produced only a few thousand in English. This disparity highlights how corpus construction is not a neutral operation but one conditioned by the ecology of the press itself. Differences in scale partly reflect the smaller size and narrower remit of the English-language press, but they also underscore divergences in editorial agendas and intended readerships. In other words, the “same” query in two press environments generates corpora of unequal weight and composition, reminding us that corpus building always embeds the specificities of the sources it draws from.

Before the 1880s, articles mentioning shopkeepers appeared only sporadically, rarely more than a handful in any given year (Fig. 3). From the 1890s, the number began to rise more consistently, and the upward trend accelerated in the early twentieth century. The 1910s and especially the 1920s marked the period of greatest visibility. The curve in Figure 4 reached its peak in the early 1920s at over 200 articles in a single year. Shopkeepers thus became a regular feature of coverage during the treaty-port boom years, when the foreign press reported extensively on urban commerce, disputes, and daily life. After this high point, the curve dropped noticeably. By the 1930s and 1940s, the number of mentions fell off sharply, reflecting both geopolitical turmoil and the shrinking footprint of the English-language press.

Figure 3. Distribution of Articles in the English-language Press: Total Number of Articles per Year (1833-1953)

When expressed as a proportion of the total output, the trajectory looks quite different (Fig. 4). In the mid-nineteenth century, when the foreign press produced relatively few articles, mentions of shopkeepers occasionally made up one or even two percent of all published pieces. These early spikes reflect the weight that small-scale commerce could have in a still-limited news environment. As the press expanded in the later nineteenth century, the proportion stabilized at a lower level, generally below half a percent. Even during the 1920s peak in absolute terms, the proportion of articles mentioning shopkeepers remained very modest. From the early twentieth century onward, the share of coverage devoted to shopkeepers steadily declined and flattened out, rarely rising above a fraction of a percent. By the 1930s and 1940s, shopkeepers had become marginal in proportional terms, their visibility dwarfed by political, international, and wartime reporting.

Figure 4. Distribution of Articles in the English-language Press: Proportion of all Articles per Year (1833-1953)

In the mid-nineteenth century, mentions of shopkeepers were scattered mainly across publications such as the Chinese Repository and the North China Herald (Fig. 5). From the 1890s onward, however, the North China Herald emerged as the dominant voice, consistently producing the majority of articles mentioning shopkeepers. Its weight derived from its distinctive orientation: unlike more specialized or missionary journals such as the Chinese Recorder, the Herald was geared toward Shanghai’s local life. As the weekly edition of the North China Daily News, it offered a digest that recorded a wide range of urban events, from commercial disputes to civic affairs, in a way that other English-language outlets rarely matched. Other titles, such as the Shanghai Times, contributed episodically but at a much lower volume. By 1900, the field had broadened slightly, with publications like the Peking Gazette and the China Weekly Review beginning to register. Still, the North China Herald remained the backbone of English-language reporting on commerce and shopkeeping, precisely because its editorial mission was so closely tied to the rhythms of Shanghai’s public and commercial life.

Figure 5. Distribution of Articles by Newspaper (English-language)

The two perspectives reveal a striking paradox similar to what we saw in the Shenbao, though on a different scale. In absolute numbers, references to shopkeepers in the English-language press grew steadily, peaking during Shanghai’s golden age as an international metropolis in the 1920s. But in relative terms, shopkeepers lost ground almost continuously from the late nineteenth century onward. Their early prominence was not sustained, and their proportional share of the news agenda declined even as their absolute visibility rose. By the interwar years, foreign readers encountered more stories about shopkeepers than ever before, yet these stories represented only a small slice of an ever-expanding and more diversified press landscape.

Conclusion

The exercise of building a corpus of shopkeepers from the Shenbao and the English-language press demonstrates that such a body of texts is never a transparent window onto the past. It is both an analytical tool and a historical artifact in its own right. On the one hand, the corpus provides scale and scope: tens of thousands of articles can be aggregated, counted, and compared, revealing long-term trajectories in the visibility of shopkeepers across seven decades. On the other hand, the very shape of the corpus is the product of choices that reflect terminological definitions, query structures, and the editorial priorities of the newspapers themselves. What is captured highlights not only the historical presence of shopkeepers but also the thematic emphases, silences, and constraints that structured the press.

Seen from this angle, the corpus must be treated not simply as raw data to be mined, but as a layered construction that encodes multiple histories: of commerce and urban life, of journalistic practice, of censorship and political upheaval, and of our own methodological interventions. Recognizing the corpus as an analytical object in its own right allows us to move beyond a purely quantitative reading of frequencies and proportions. It reminds us that every spike, trough, or absence in the data is meaningful only when situated within these intersecting histories. Corpus building thus does more than generate material for further analysis; it offers a critical vantage point on the production, circulation, and preservation of public discourse in modern Shanghai.

Methodological Note

The shopkeeper corpus was constructed through a workflow that combined computational query expansion, large-scale retrieval, data cleaning, and statistical analysis. The initial entry point was the lexical field of “shopkeeper” in both Chinese and English. For the Shenbao, the initial term 店主 (dianzhu, “shop owner”) was expanded into a larger set of sixteen terms using the HistText Word Embeddings function, which identifies semantically related vocabulary based on distributional similarity. This allowed the capture of historical variants and context-specific synonyms. In the English-language press, queries were more standardized but still accounted for orthographic variants and phrase-level expressions.

Using the search_documents() function of the histtext R package, I queried the revised Shenbao digital corpus for all terms simultaneously, while also generating term-specific subsets for comparative purposes. For the English-language press, the same procedure was applied to digitized collections of the North China Herald, North China Daily News, and related publications. The combined search yielded: 69,851 articles in Chinese (after deduplication) and 5,145 articles in English. Deduplication ensured that articles containing multiple terms were counted once.

Using R, I aggregated both corpora by year to generate time series plots of article counts (absolute frequency) and relative coverage (percentage of total newspaper output). For Chinese, I further examined the temporal distribution of each term, generating multi-line visualizations of lexical trajectories. Comparative plots aligned Shenbao and English-language outputs to highlight cross-press divergences. Visualizations were produced using the ggplot library in R.

Statement on AI-Assisted Development

The construction and analysis of the shopkeeper corpus were supported at various stages by the use of large language models (LLMs). Their role was not to replace domain expertise but to augment and accelerate specific tasks that are otherwise highly time-consuming. During the data processing stages in R, LLMs were employed as technical aides to:

  • troubleshoot scripts,
  • suggest efficient ways of structuring loops and filters,
  • check regular expressions used in false positive removal,
  • and generate draft templates for visualization.

All scripts were then verified, executed, and refined manually within the R environment. The outputs relied on reproducible code rather than on LLM-generated results.

At the interpretive stage, LLMs were used to synthesize and structure intermediate notes, helping to turn fragmented observations (e.g., from temporal series, proportional analyses, or concordances) into coherent analytical summaries. Finally, LLMs were used selectively for stylistic refinement to smooth transitions, harmonize terminology, and maintain consistency of tone across sections. All substantive arguments, historical claims, and interpretations remain the result of close reading, computational analysis, and the author’s historical expertise. This statement makes explicit that AI tools were assistants, not authorities: they facilitated coding, drafting, and summarization, but the design of the corpus, the critical evaluation of sources, and the interpretation of results rest firmly on human scholarly judgment.


[1] Cécile Armand, Between Public and Market: A Spatial History of Advertising in Modern Shanghai (1905–1949) (Berlin:De Gruyter, 2025).


OpenEdition suggests that you cite this post as follows:
Christian Henriot (October 7, 2025). Behind the Counter: When Shopkeepers Made (and Lost) the News. Blog Virtual Shanghai. Retrieved November 12, 2025 from https://doi.org/10.58079/14vdw


You may also like...

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.