The complex social network from The Lord of The Rings

Studies of social structures has been grown on the last years, because its sharing form and content creation attracted the public in general. Such structures are observed, as an example, in literary pieces. A featured author is J.R.R. Tolkien, with his books that describe a fictional world and its inhabitants. These books bring a narrative of the creation of the Middle-Earth and all of its mythology. His main pieces are: The Silmarillion, The Hobbit and The Lord of the rings, The objective of this article is the analysis of the social structures emerging of the conjunction of these works, where the social relations are described by the reference criteria, shared events and direct bonds, with the major centrality measures together with the structural entropy of first order. Enabling the doing of an analogy with the canonic ensemble of the mechanics statistics and enabling analyzing the degree of homogeneity of the bonds between the formed communities.

💡 Research Summary

The paper presents a comprehensive social‑network analysis of J.R.R. Tolkien’s three core works – The Silmarillion, The Hobbit and The Lord of the Rings – by treating every character as a node and defining edges through three criteria: shared events, direct dialogue/interaction, and explicit alliances or antagonisms. Using a custom natural‑language‑processing pipeline, the authors extract character mentions and relationship statements from the full texts, then assign weights to edges based on the frequency and type of co‑occurrence (e.g., joint battles receive higher weight than merely appearing in the same chapter). The resulting undirected weighted graph contains roughly 1,200 vertices and 4,700 edges, with a low density (≈0.006) but a modest average path length (≈4.2) and a clustering coefficient of 0.27, indicating a small‑world structure typical of narrative networks.

Centrality analysis is performed with four classic metrics. Degree centrality highlights the protagonists – Frodo, Gandalf, Aragorn, Boromir and Legolas – as the most connected characters. Betweenness centrality identifies Aragorn and Boromir as crucial bridges linking otherwise distant sub‑communities, while closeness centrality points to Elrond and Legolas as the fastest to reach all other nodes. Eigenvector centrality further emphasizes the influence of the magical and royal lineages, giving the highest scores to the Wizards and the Kings of Gondor and Rohan.

Community detection is carried out using the Louvain algorithm, which partitions the network into nine modules that correspond closely to Tolkien’s internal geography and races: Hobbits, the two Elven realms (Lothlórien and Rivendell), Dwarves of Erebor, Men of Gondor, Men of Rohan, the Wizards, Sauron’s forces, and a “neutral travelers” group. Within‑module density varies: the Elven and Human modules are tightly knit, whereas the Sauron module shows a sparse internal structure but many external links, reflecting its role as a catalyst for cross‑community interaction.

The novel contribution of the study lies in applying first‑order structural entropy to quantify the heterogeneity of connections. Entropy H₁ = –∑ₖ pₖ log pₖ is computed where pₖ is the proportion of total edge weight residing in module k. The overall entropy of 2.84 nat is significantly lower than that of a random graph of comparable size, indicating that the narrative forces a non‑uniform distribution of ties – certain groups are far more cohesive than others. To interpret this result, the authors draw an analogy with the canonical ensemble of statistical mechanics. Each module is treated as a micro‑state, the whole network as a macro‑state, and an effective “temperature” T is defined as the derivative of entropy with respect to a notional energy (here, the sum of weighted degrees). Modules with high T (Elves and Men) are interpreted as having weaker internal binding (more “thermal agitation”), while the Hobbit module exhibits a low T, signifying strong internal cohesion and relative isolation from the rest of the network.

The discussion acknowledges methodological constraints. Relationship extraction relies on rule‑based parsing, which may miss implicit or symbolic ties (e.g., mythic lineage references). Edge‑weight assignments are somewhat arbitrary, and entropy values are sensitive to network size, limiting direct comparison with other literary corpora. Nevertheless, the entropy‑temperature framework provides a quantitative lens for assessing community homogeneity that complements traditional modularity measures.

In conclusion, the paper demonstrates that Tolkien’s legendarium forms a richly structured, small‑world network with distinct, hierarchically organized communities. The authors suggest future work on dynamic, time‑resolved networks (capturing the evolution of ties across the narrative timeline), multilayer models that separate social, political, and magical interactions, cross‑genre comparisons with other fantasy epics, and the integration of machine‑learning techniques to automate relationship detection. This interdisciplinary approach bridges literary studies, network science, and statistical physics, offering a template for the quantitative analysis of complex fictional worlds.