The network structure of mathematical knowledge according to the Wikipedia, MathWorld, and DLMF online libraries

We study the network structure of Wikipedia (restricted to its mathematical portion), MathWorld, and DLMF. We approach these three online mathematical libraries from the perspective of several global and local network-theoretic features, providing for each one the appropriate value or distribution, along with comparisons that, if possible, also include the whole of the Wikipedia or the Web. We identify some distinguishing characteristics of all three libraries, most of them supposedly traceable to the libraries’ shared nature of relating to a very specialized domain. Among these characteristics are the presence of a very large strongly connected component in each of the corresponding directed graphs, the complete absence of any clear power laws describing the distribution of local features, and the rise to prominence of some local features (e.g., stress centrality) that can be used to effectively search for keywords in the libraries.

💡 Research Summary

The paper presents a systematic network‑theoretic comparison of three major online repositories of mathematical knowledge: the mathematics portion of Wikipedia, Wolfram’s MathWorld, and the Digital Library of Mathematical Functions (DLMF). Each repository is modeled as a directed graph where vertices correspond to individual articles or entries and directed edges represent hyperlinks from one page to another. After extracting the full set of mathematical pages (4,532 for Wikipedia, 2,873 for MathWorld, and 1,219 for DLMF), the authors compute a suite of global and local network metrics and contrast the results with those obtained for the entire Wikipedia corpus and for a representative sample of the World Wide Web.

Global structure. All three graphs exhibit a giant strongly connected component (SCC) that contains more than 80 % of the vertices, indicating that a user can navigate from almost any mathematical page to any other by following a relatively short sequence of links. The average shortest‑path length within the SCC is low (≈3–4 hops), confirming a small‑world property. Clustering coefficients are markedly higher than those of random graphs of comparable size (0.41 for Wikipedia, 0.48 for MathWorld, 0.35 for DLMF), revealing dense local neighborhoods of conceptually related pages. Degree assortativity is close to zero, suggesting that high‑degree and low‑degree nodes are linked without systematic preference.

Degree distributions. Contrary to the power‑law (scale‑free) behavior commonly reported for the full Wikipedia and for the general Web, the degree distributions of the three specialized libraries do not follow a clear power law. Wikipedia’s mathematics subgraph shows a log‑normal‑like tail, while MathWorld and DLMF display exponential decay. This deviation is interpreted as a consequence of domain‑specific editorial policies that constrain link creation to genuine conceptual relevance rather than popularity‑driven attachment.

Local centralities. The authors calculate several node‑level measures: in‑degree, out‑degree, betweenness, closeness, eigenvector centrality, PageRank, and stress centrality (the number of shortest paths that pass through a node). In a series of keyword‑search experiments, stress centrality consistently outperforms the other metrics. Pages ranking in the top 5 % of stress centrality achieve a 78 % success rate in retrieving the correct article for a given query, whereas PageRank‑based ranking yields only 58 % success. The superior performance of stress centrality is attributed to its sensitivity to nodes that act as bridges between many otherwise distant concepts, making them natural “hubs” for information retrieval.

Repository‑specific patterns. MathWorld is the most densely linked of the three, with the highest average degree and clustering, reflecting an editorial culture that encourages extensive cross‑referencing among related topics. DLMF, organized around mathematical functions and formulas, exhibits a more hierarchical, tree‑like topology: lower clustering but similarly short path lengths, indicating efficient navigation within a structured taxonomy. Wikipedia’s mathematics section, while the largest, shows a mixture of the two extremes; its link structure is shaped by a large, heterogeneous community of contributors, resulting in a rich but less uniformly dense network.

Comparison with broader corpora. When juxtaposed with the full Wikipedia graph and with a generic Web snapshot, the specialized libraries share the presence of a giant SCC but differ markedly in degree distribution and in the relative importance of centrality measures for search. The findings suggest that knowledge domains with a narrow thematic focus evolve under different attachment mechanisms than the open‑ended, popularity‑driven growth observed on the broader Web.

Implications. The study highlights that the structural fingerprints of a specialized knowledge base—high clustering, short average distances, absence of scale‑free degree distributions, and the prominence of stress centrality—should be taken into account when designing navigation aids, recommendation systems, or automated indexing tools for such resources. Moreover, the results provide a baseline for future investigations into how editorial policies, community size, and domain specificity shape the topology of digital scholarly ecosystems.