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

What’s under the surface? Building and exploring latent connections in ManMax data | by Richard Hadden & Marcella Tambuscio


This blog post is in two parts: the first explores our continued efforts to gather data for the Managing Maximilian project’s prosopographical database. The second showcases the potential for network analysis techniques on highly interconnected data, carried out on the ManMax data by Marcella Tambuscio.

 

Data Development

In order to enable the kinds of network analysis we are interested in, it is necessary to have a lot of data, ideally creating a highly interconnected network. So far, the ManMax project has succeeded in manually collecting and entering over 3000 Factoids (for more information on the Factoid model used by ManMax, see our previous blog post: https://manmax.hypotheses.org/2994).

However — and often because it is the most interesting aspect for historians! — these Factoids are largely concerned with single events and interactions. While any analysis that missed these elements would indeed be lacking, it creates another issue: the focus of each subproject on their own relevant persons and kinds of events elevates the importance of such interactions and produces a disconnected network of “sub-project-specific” groups.

It would be highly improbable if the only persons that, for instance, a writer (focused on by the Writing Maximilian subproject) only interacted with other writers and publishers (the subproject’s area of interest), and didn’t interact with his wife! However, specific evidence of interactions with the wife are likely to be thin on the ground. What we lack, in other words, is information on more latent relations — family membership and relations; roles in organisations and households — that underpin and contextualise the specific interactions. Consequently, we need to add as many of these scarce indicators for relations as possible, for instance by drawing on existing data sets.

Multi-layered network showing connections between family relationships and interactions

 

Mining the Regesta Imperii Index

One key source for ManMax was the index compiled through tenacious work by Angelika Schuh for the Regesta Imperii XIV. Ausgewählte Regesten des Kaiserreiches unter Maximilian I. 1493–1519 (Register der Personen- und Ortsnamen 1496–1498, 1993 and 1996), the raw data for which she very kindly provided to the project. (In the project, we call it the “Schuh Index”.) It has already been an invaluable resource to ManMax. It comprises a detailed list of names, families, organisations and places from the period of Maximilian, which we extracted and (thanks to the phenomenally hard work of Sonja Lessacher) manually classified. This index served as an initial data ingest for ManMax, allowing us to avoid duplication and the manual effort of adding a whole coterie of “likely suspects”.

A page from the “Schuh Index”

The utility of the Schuh Index does not stop at being merely a useful list, however. Its original structure — it was originally a print index — of lemmata and sub-lemmata provides useful, albeit unclassified, information (while it is obvious to a human reader that the connection between the lemma Aachen and sublemma Hans von Reutlingen, Goldschmied zu Aachen is one of person-in-a-location, and that of lemma von Adelsheim to sublemma Zeisolf von Adelsheim denotes a family relationship, this is not obvious to a computer!) But, thanks to Sonja’s work classifying the entries of the index, we are able to make a number of deductions: for example, that sublemmata of type Personcontained in a lemma of type Family denotes a family relationship; similarly for persons in locations.

While these connections are rather prosaic and imprecise (“family membership” is rather more vague than actual parent-child/sibling relations), the sheer amount of information obtained (5496 Factoids detailing such relations) begins to illustrate latent ties between persons, creating a familial substrate for our network.

We can even take this a step further and look at information contained in the index labels themselves. Using simple techniques such as Regular Expression matching, we can identify key phrases — “Ehefrau von”, “Sohn von” — and extract name following the phrase. From “Ursula Aicher, Ehefrau von Stefan Ranshofer, aus Wels”, for example, we extract the name “Stefan Ranshofer”, which we then look up in the index and create the appropriate Marriage factoid.

 

Towards Large Language Models

Elementary pattern matching will, alas, only take us so far. The labels of the Schuh Index contain a wealth of further information, which is not readily parseable by such techniques. The grammatical distinction between commas separating a list of roles, as in “Antoine de Crenel/Coinel [Cribellus], Titularbischof von Betlehem (1501-1512), Abt des Klosters L’Etoile bei Poitiers”, and commas denoting a relative clause, as in “Sibilla N. N., Ehefrau des “jungen” Aichorn von Hall, Dienerin von Bartholomaüs Käsler, der Witwe des Hans Käsler sowie von Dorothea Schrofenstein” is not obvious enough without understanding the sense. (Antoine de Crenel has multiple roles; Sibilia certainly was not the widow of Hans Käsler). For this, we hope to employ the Large Language Model approach already utilised by Suzana Sagadin and presented in its preliminary forms at the “Generative KI in der mediävistischen Grundlagenforschung” conference in Saarbrücken a year ago. The publication on this will be ready soon, so stay tuned!


Now, we turn to Marcella Tambuscio’s most recent network analysis, exploring the possibilities of the dataset.

 

Bridging people, actions and structures: exploring typed networks at Maximilian’s Court

Prosopography traditionally studies historical interactions among individuals or groups, and network analysis is a method that provides mathematical tools, such as centrality and clustering measures, to explore these connections. What happens when we treat historical interactions not just as basic connections, but as meaningful actions?

In the Digitising Maximilian sub-project of ManMax, we experimented with network models that go beyond traditional co-occurrence or binary ties. Rather, we look at what people actually did together: giving a gift, commissioning a text, printing, attending a performance, issuing an order… (We have over sixty types of interaction, in fact). Each of these actions leaves a trace in historical sources: in our model, it becomes a typed edge in a growing network of individuals surrounding Emperor Maximilian I. Instead of simply linking people who appear in the same document, we classify edges by type, creating a semantic multigraph: individuals can be connected multiple times in different ways. This lets us ask questions like:

  • Are people who exchange money also involved in other types of interactions?
  • Are communities of actors built around functions (printing, diplomacy) or hierarchy?
  • Who acts as a broker between groups? And how diverse are these roles?

Here’s a snapshot of the network, built from data already integrated into our digital prosopographical model. For clarity, we’ve temporarily removed Maximilian because he’s involved in almost everything, and acts as a “superhub” that can obscure the structure of everyone else’s relationships.

Interaction network extracted at July from APIS with different visualisations: the automatic community detection algorithm highlights the structural components of the networks (on the left) while using different colors for different types of links exhibits semantic structure of interactions

We applied community detection algorithms to identify groups in the network based on connection density. Interestingly, some of these communities align with functional domains, while other exhibit a variegate composition. But we wanted to go further: to combine network structure with semantic information. So we looked at two things for each person:

  • How many of their connections are across communities (inter-community ratio);
  • How diverse their actions are, measured using a normalized entropy score

By plotting individuals in this space, we can classify them into four interpretive roles, as listed in the following scatterplot:

People (involved in more than 10 connections in current version of APIS) positioned according to their semantic diversity (normalized entropy) on the x-axis and inter-community edge ratio on the y-axis. This two-dimensional representation allows for the classification of individuals into four interpretive categories combining the diversity of their interactions and their position in the network.

This classification isn’t just a technical curiosity: it points us toward people who deserve a closer look. Take Johannes Cuspinian: humanist, diplomat, curator, and poet. He appears as a cross-community generalist, and the network confirms what historical scholarship already suggests: he operated across many domains in Maximilian’s court. Meanwhile, Matthäus Lang and Jakob Wimpfeling are cross-community specialists, involved mostly in one kind of action, but connecting different groups. Others remain mostly within one group, but take on a wide range of roles, like Bianca Maria Sforza, or just one, as with Johannes Stabius.

Typed networks help us to go beyond structure: they offer a way to see not just who is connected to whom, but how and what that means. By combining formal analysis with historical interpretation, we can uncover roles that aren’t always obvious from the sources themselves. This is just the beginning. Our next steps include extending the network with more sources and types, incorporating the temporal dimension to explore change and evolution, and connecting these roles with deeper archival readings.


OpenEdition suggests that you cite this post as follows:
marcellatambuscio (September 9, 2025). What’s under the surface? Building and exploring latent connections in ManMax data | by Richard Hadden & Marcella Tambuscio. Managing Maximilian (1493-1519) - A blog on Persona, Politics, and Personnel through the Lens of Digital Prosopography at the time of Maximilian of Habsburg. Retrieved February 7, 2026 from https://doi.org/10.58079/14m8v


You may also like...