This project uses computational text analysis, particularly classification models, to examine the uneven circulation of Global South novels across international book markets. By combining digital humanities methods with literary-sociological approaches to the global literary supply-chain, it demonstrates how textual patterns and distribution infrastructures jointly shape the global movement of postcolonial novels like those written by Malaysian and Zimbabwean authors. The project operationalizes scholarship by postcolonial book historians like Graham Huggan (2006), Sarah Brouillette (2007) and Ponzanesi (2014).

In this project, I propose global literature to refer to literatures that enter the global book-supply chain and circulate beyond their region of origin. This framework distinguishes global literature from world literature—which may have a more evaluative or canonical sense—as an infrastructural category that identifies texts that move through repeatable cross-market distribution mechanisms. Rather than assuming that texts travel because of the quality of its writing, the term considers the infrastructural conditions—the licensing channels, retail systems, and market intermediaries—that determine whether a book moves into multiple markets or remains restricted to its original point of production. Under this definition, circulation is not treated as a sign of aesthetic value or international appeal, but as an outcome of supply-chain positioning.

The project draws on an original corpus of 216 Anglophone novels published between 2000 and 2025 by Malaysian and Zimbabwean authors. The period of 2000-2025 is chosen primarily to create a snapshot of the most current practices so as to inform future actions—such as policy recommendations—while also capturing enough data to make claims about the industry. The post-2000 period is also marked, in the case of Zimbabwe, by a sharp economic downturn and increasingly organized opposition to the incumbent Zimbabwe African National Union – Patriotic Front (ZANU-PF) government.

Novels by Malaysian and Zimbabwean authors were selected for this study’s sample as much for the countries’ similarities as for their differences. They are similar in that both represent Anglophone countries that were previously colonized by Britain, but they exist in very different regional contexts—Zimbabwe is located within a larger and more cohesive African framework (whether in terms of “African literature” as a coherent category or in terms of continental treaties and organizations that enable better regional movement and collaboration) whereas Malaysia is located within a smaller and more fragmented region of Southeast Asia that has a variety of colonial histories and competitive national and regional language markets. Although Malaysia and Zimbabwe’s national profiles are dissimilar—Malaysia has more than twice the population and more than 5 times the GDP per capita —their outputs when it comes to novels are comparable, with Zimbabwean novels published in the last 25 years outnumbering Malaysia (116 vs 100). The number of novels published by Malaysian authors grew more recently, with Zimbabwe displaying a steadier and earlier output belying its longer held status as a literary hub (with, for example, the renowned Zimbabwe International Book Fair that began in 1983 and had over a hundred stands every year). An additional similarity between Malaysia and Zimbabwe is their connection with a stronger, neighboring book market—Singapore and South Africa. Granted, data taken from a sample of two countries would inevitably be incomplete in analyzing a global system, but Malaysia and Zimbabwe’s regional differences allow me to illuminate aspects of global distribution through the patterns that they do share. The criteria for titles included in the dataset are as inclusive as possible. Any author that was born in Malaysia or Zimbabwe and/or claims a connection to Malaysia or Zimbabwe is included. The broad definition of national belonging includes diasporic authors, which characterizes many of the authors in the dataset. Only novels originally written in English are included, which is especially necessary in this case to distinguish between territorial or world licensing rights and translation rights.

Titles were drawn from every avenue, ranging from publishers’ catalogues, personal conversations, and digital archives for book metadata like Reading Zimbabwe. The resulting dataset contains 216 unique titles, 100 by Malaysian authors and 116 by Zimbabwean authors. For each title, I recorded detailed bibliographic and publication metadata, including publisher, imprint, parent company, place of publication, and year of release. To examine how these novels move through book-supply systems, I collected availability data from the online or bookstore platforms located in four retail markets—Amazon in the United States, Amazon in the United Kingdom, Kinokuniya bookstore in Malaysia, and takealot in South Africa—using consistent criteria for new, purchasable physical copies with reasonable pricing and delivery timeframes. As Zimbabweans often purchase goods from South Africa, with many hired “runners” bringing in items across the border, South Africa is included as the closest proxy for the Zimbabwean market, which does not currently have online platforms with a reliable catalogue of purchasable books. Treating availability as a measurable indicator of circulation converts “circulation” into a reproducible infrastructural variable suitable for computational modeling. All 216 novels are currently being digitized through a combination of destructive and overhead scanning and OCR to prepare them for downstream text-analysis workflows.

Early analysis of the metadata and availability data already provides important insights. Figure 1 visualizes availability patterns across the four markets. The results reveal a notable asymmetry: UK and US readers have significantly greater access to Malaysian and Zimbabwean novels than readers in the novels’ countries of origin. Malaysia’s import-heavy retail ecosystem ironically affords Malaysian novels comparatively stable local distribution. Zimbabwean novels, however, remain inconsistently available within the South African/Zimbabwean regional market. This pattern is corroborated by Table 1, which shows that many more Zimbabwean novels are only available in the UK or US and not in South Africa, while 28 are not available at all.

A second analysis examines the relationship between publisher and licensor location and circulation. Figure 2 compares novels published and/or licensed by multinational or foreign presses with those published solely by local presses. Across both Malaysian and Zimbabwean corpora, foreign-published novels display dramatically wider circulation, while local-only publications exhibit highly uneven distribution. Malaysia produces relatively similar trends for foreign and locally published books, wherein locally published books continue to be significantly better distributed in the UK market, especially, rather than in the local market. In Zimbabwe, by contrast, locally published books show somewhat better circulation within the regional market.

These analyses are only preliminary, with many more variables to be explored. It also necessarily follows that these unequal distributional systems and their market pressures could meaningfully affect what books can be written and read. This raises the question: do globally circulating novels share discernible textual or stylistic features that differentiate them from more locally circulating works? And if so, can computational methods surface these features within a framework attentive to the pressures exerted by distributional infrastructures?

To answer this, the project employs a two-stage modeling strategy integrating unsupervised and supervised methods, especially supervised classification models. The project follows the footsteps of other computational literary study scholarship that has used or advocated for classification models (see, for example, Piper 2018, Long and So 2016, or more recently, Bammam et al. 2024).

First, I apply unsupervised methods run separately on the locally circulating and the “not available anywhere” corpus and the globally circulating corpus. This stage is exploratory, not to infer causality but to map internal variation and generate grounded hypotheses. It identifies the thematic and stylistic features characteristic of each circulation regime. Because the corpus consists of 216 novels with a much smaller segment of local or non-circulating texts, unsupervised modeling will be conducted on text segments (e.g., chapters or fixed-length windows) rather than on whole books.First, I apply unsupervised methods run separately on the locally circulating and the “not available anywhere” corpus and the globally circulating corpus. This stage is exploratory, not to infer causality but to map internal variation and generate grounded hypotheses. It identifies the thematic and stylistic features characteristic of each circulation regime. Because the corpus consists of 216 novels with a much smaller segment of local or non-circulating texts, unsupervised modeling will be conducted on text segments (e.g., chapters or fixed-length windows) rather than on whole books.

Second, I introduce supervised classification models to evaluate which textual features predictively distinguish global from local circulation. Availability outcomes across the four markets and publisher-scope categories (foreign vs. local only) serve as labels for training. Features would be derived from the exploratory stage. This confirmatory stage provides interpretable evidence of how circulation and textual patterns correlate.

The two stages illuminate whether and how textual features align with circulation patterns shaped by the global book-supply chain. The project ultimately demonstrates how computational approaches, paired with sociological analysis of literary circulation, can offer new insights into how distributional forces are shaping contemporary global literature.