Trismegistos Networks

Perhaps you’ve noticed, perhaps you haven’t. On some Trismegistos pages, you can discover delightfully stringy contraptions called networks. In technical lingo, these are also known as spaghetti monsters. And no, we’re not making that up. They really are.

What do spaghetti monsters do, you ask? Networks are the perfect tool to study relations. The most straightforward type are social relations of course, interactions between individuals. But when you think of it, almost everything is related to something else in our world, so if you’re creative, you can probably come up with a lot of things to take a look at from a network perspective. Since Trismegistos has a lot of texts, and those texts mention a lot of people, a network of people appearing in the same text is easily concocted. The same criterion can be used to create a network of toponyms. Names can also be noodlized: in the Trismegistos People section, each name page visualizes how it is linked to other names by means of genealogical relationships. Our editor database provided us with plenty of data to create co-authorship networks. And we’re just getting warmed up here. With all the information we’ve got in Trismegistos, there’s much more where that came from!

On this page, you can browse and sample our finest selection in the menu, brush up your basic network theory (and pick up some tips on how get the most out of these online visualizations), and if you want to have a go at making networks yourself, we can help you get started as well.

If you’re interested in spaghetti monsters and think they can help you with your research, make sure to check out our historicaldataninjas blog as well!

What is a network?

Networks are ways to map connections between elements. Many networks concern relationships between people, but anything can be the subject of a network, as long as there's a relation between the elements. Each person, or settlement, or name, etc., is represented by a node (the dots), and each connection is shown as an edge (or link or tie: the lines between the dots).

The nature of the connection can be chosen freely. A network can be undirected, i.e. all ties are reciprocal, or directed if there is a certain "information flow" from one node to the other (this type of connection is indicated by an arrow at the end of the edges).

There are many network measures to describe the features of a network, either as a whole, as well as those of individual nodes and ties: betweenness centrality, eigenvector, density, modularity, ... In our networks, we have limited the information to properties pertaining to certain nodes.

In principle the shape of the network is irrelevant: as long as nodes are connected, they can be placed anywhere. But as humans have difficulty seeing the jumble of edges between the thousands of nodes in the large networks we create, programs like Gephi can adapt the visual representation of the network according to specific layout algoritms. To bring out the structure in our networks more clearly, we use layout algorithms such as 'ForceAtlas2' or 'YifanHu'.

Degree centrality (or simply 'degree'): this is a basic centrality measure that counts the number of ties a node has. Regarding people, this can be seen as a popularity measure, since having many links implies that that person has many acquaintances. In a directed network, a distinction is made between in-degree and out-degree:

In-degree: these are the number of edges pointing to a certain node. In a directed network, the in-degree is used to measure popularity, since it is an indication of how many sources refer to that particular node.
0ut-degree: these are the number of edges starting from a certain node. This can be interpreted as the node's 'outreach'.
Weighted degree: a connection between two entities can be registered more than once (see above), and this edge weight can be taken into account when calculating degree centralities.

Betweenness centrality: this centrality measure does not count the number of direct links a node has, but rather how often it lies on the shortest path between two other nodes. Nodes with high betweenness often play a broker role in the network. If node A and node C are both connected to B, but not directly to each other, B is crucial to get information from A to C and vice versa.

Geodesic distance and eccentricity: the geodesic distance is the number of steps it takes to get from one node to another, using the shortest path (as in the famous 'six degrees of separation'). A geodesic distance of 1 means that two nodes are directly linked. A geodesic distance of 5 means that node A has to go through minimum 4 other nodes to get to node B. Eccentricity then measures how far a node is from the furthest other connected node, i.e. its largest geodesic distance.

Closeness centrality: again, this centrality measure does not focus on a node's direct links, but rather at how connected it is to the entire network. Closeness is based on the inverse of the distance of a node to all other connected nodes, so here the bottom line is: the lower the result, the better. Node A, despite the fact that it is only connected to node B, can still have a high closeness centrality if node B has many connections to other (central) nodes in the network.

Modularity class: community detection is another network analysis feature. We use the Louvain algorithm in Gephi to detect sub-units in large networks, i.e. highly interconnected groups of nodes with only sparse connections to other groups in the network. By looking at the links between the nodes, the computer assigns them to 'classes' or 'communities', which are indicated by a random number. We can then look for meaningful characteristics that create such a structure. In the geographical networks (network 7 & 8), for example, the communities largely coincide with Egypt's administrative district.

Make networks yourself

If you are an absolute beginner in network analysis, but are tempted to try it yourself, we recommend our blog Six degrees of spaghetti monsters. It is full of information about network analysis, and contains many practical tips on how to get started, many of them using the data in Trismegistos.

The nice thing about Trismegistos is that, thanks to its relational structure linking the people/names/places (everything, basically) databases to the text database, we have a basic relation to start with. This is called a two-mode network, since it consists of two different entities: people appearing in texts. But most of the time you’ll be interested in one-mode networks: people linked to people, for example. This information can’t be extracted directly from Trismegistos, or many other databases for that matter.

There are all kinds of ways to convert two-mode networks into one-mode networks, but this requires knowledge of specialized software such as R or UCINET. Yanne takes care of it with a flick of her wand of course, but in all fairness, it did take her almost two years to manage this nifty craft. So to help people out, we’ve created a Filemaker tool that takes care of that for you. We have called it the tomator, for not-so-obvious reasons. But it does work, and the good news is: all you have to do is push a magic button. Check it out here.