Digital Humanities Text Mining for Postcolonial Literary Networks

Thesis Abstract

This study addresses the fragmentation of postcolonial literary networks in the digital age, where traditional scholarly discourse has often overlooked transregional interactions and intertextual flows across colonial and postcolonial archives. The problem is twofold first, limited coding schemes and sparse metadata impede scalable mapping of literary influence; second, existing digital humanities approaches frequently underutilize network theory to capture latent connections among authors, genres, and publishing ecologies. The aim is to develop a scalable text-mining and network-analytic framework that uncovers postcolonial literary networks through machine-assisted data extraction and theory-driven interpretation. Specific objectives are (1) to compile a corpus of 1,200 digitized literary texts spanning Africa, the Caribbean, South Asia, and the metropole, published between 1850 and 2000; (2) to construct a multilayer network incorporating co-citation, intertextual allusion, thematic similarity, and publishing-ecology links; (3) to apply topic modeling, stylometric analysis, and named-entity recognition to identify latent motifs and authorial communities; (4) to test hypotheses derived from postcolonial theory using network centrality, modularity, and assortativity measures; and (5) to formulate a replicable analytical pipeline for DH-based investigations of postcolonial literary systems. The methodology integrates a mixed-methods design anchored in digital humanities and network theory. The population consists of digitized literary corpora from major archives (British Library, Bibliothèque nationale de France, University of Chicago Library) and regional repositories. A stratified random sample of 1,200 texts is drawn to balance geographic representation and publication period. Data collection instruments include (a) a customized text-mining pipeline using Python with spaCy for named-entity recognition and Gensim for topic modeling, (b) metadata augmentation through library catalog records and OCR quality checks, and (c) a coding schema for intertextual and thematic linkages informed by Postcolonial Theory (e.g., Edward Said, Homi Bhabha) and Network Theory (e.g., Freeman, Barabási–Albert). Validity and reliability are enhanced through triangulation across three data streams (textual content, metadata, and citation/intertextual proxies) and inter-coder reliability tests on a 10% subset of annotated links. Data analysis proceeds in multiple stages (i) preprocessing and normalization of texts, (ii) unsupervised topic modeling (LDA and BERTopic) to extract themes, (iii) stylometric profiling (word n-grams, function-word usage) to infer authorial similarity, (iv) construction of a multiplex network with layers for co-citation, intertextual allusions, thematic affinity, and publishing-ecosystem proximity, (v) network analytics including degree centrality, betweenness, eigenvector centrality, modularity-based community detection, and assortativity by region and era, and (vi) regression-based significance testing to assess the relationship between centrality measures and thematic convergence. The analytical framework draws on postcolonial theory to interpret the significance of discovered communities and cross-regional linkages, and on social-network theory to explain emergent structures as evidence of transnational literary exchange. Expected findings include the identification of previously underappreciated transregional networks linking authors across the Atlantic, Indian Ocean, and Caribbean littératures, and the demonstration that thematic convergence and intertextuality migrate along publishing ecologies and editorial networks. The study anticipates demonstrating how central nodes function as brokers of cross-cultural dialogue, and how peripheral clusters reveal latent diasporic circuits. The contribution to knowledge lies in delivering a rigorous, scalable DH pipeline that combines topic modeling, stylometry, and multiplex network analysis to illuminate postcolonial literary networks, providing empirical substantiation of theoretical claims regarding transnational influence and circulation. The study concludes with recommendations for repository metadata standardization, a publicly accessible replicable codebase, and guidelines for applying similar DH-network approaches to other literatures, with potential policy implications for digital philology curricula and grant-funding priorities in humanities data science.