What 698 Million Edges Actually Looks Like
Numbers at this scale are difficult to make concrete. Consider that the entire English Wikipedia contains roughly 6.7 million articles. Each article links to other articles, dozens of them on average. When you follow every one of those outgoing links, resolve redirects, and map each link to the pair of knowledge-graph entities it connects, you end up with approximately 698 million directed edges. That is the raw material of Vinony's Graph portal.
Rendered naively, 698 million edges would produce an unreadable tangle. The Graph portal does not render all of them simultaneously. Instead, it begins with a subset centred on whichever entity you are currently exploring, drawing the most significant connections first. Those are the connections with the highest link weights, derived from how frequently one entity's Wikipedia article links to another's. As you navigate, the graph expands and redraws around your position in the network.
The visual result is a kind of living diagram of conceptual proximity. Two entities drawn close together and connected by a thick edge have a strong mutual reference relationship in Wikipedia. Entities that appear only at the periphery of a cluster have weaker or more specialised connections to its centre. The graph does not impose an interpretation — it reflects the collective editorial judgement of millions of Wikipedia contributors about what belongs near what.
In-Degree vs Out-Degree: Why Some Entities Dominate
Every entity in the graph has two key metrics: in-degree, the number of other entities that link to it, and out-degree, the number of entities it links to. These two numbers tell very different stories about an entity's role in the knowledge graph.
High in-degree means many other Wikipedia articles reference this entity. It is a concept that turns up in discussions of many other topics. Countries, major historical periods, fundamental scientific concepts, and famous historical figures tend to have high in-degree. They are the gravitational centres around which clusters of knowledge orbit. You can explore the most interlinked concepts in the knowledge base to see which entities have accumulated the highest in-degree scores across the full graph. The results often challenge intuitions about what constitutes central knowledge.
High out-degree, by contrast, means an entity's Wikipedia article contains many links to other entities. Survey articles, disambiguation pages, and list-type entries tend to have high out-degree. They are connectors and navigators rather than gravitational centres. The distinction matters for research: a high in-degree entity is a focal concept; a high out-degree entity is a gateway.
How to Explore a Cluster: Starting with One Entity
The most effective way to use the Graph portal is to begin with an entity you understand well, observe the cluster it sits in, and then deliberately click on a neighbour that surprises you. The familiar entity is an anchor; the unexpected connection is where discovery happens.
Search for the entity by name or arrive at it from an entity page. The graph renders its immediate neighbourhood, typically the twenty or thirty entities most strongly connected to it. Take a moment to read those connections before clicking anything. Are the strongest connections the ones you expected? Where are the unexpected neighbours that do not obviously belong?
Then click one of those unexpected neighbours and watch the graph redraw around the new centre. The cluster you just left is now visible as a region in the wider graph, and the new entity's own neighbourhood is foregrounded. Continue this traversal for four or five steps and you will often find yourself in a part of the knowledge graph that feels entirely unrelated to where you started — which is the graph's way of showing you that the conceptual distance was smaller than you imagined.
Unexpected Patterns: Which Topics Are Most Connected?
Anyone who spends time navigating the Graph portal will encounter a counterintuitive pattern: some of the most connected entities in the graph are not famous people or major nations. They are concepts, abstract or categorical ideas that turn up as reference points in hundreds of thousands of different Wikipedia articles.
Calendar years rank extraordinarily high, for instance, because virtually every article about a person, event, or discovery references at least one year. Broad geographic categories (continents, major language families, large countries) appear as dense hubs because they are mentioned in the context of almost every entity that has a geographic dimension. And some scientific concepts that are fundamental to many fields simultaneously accumulate connections across disciplines in ways that more specialised ideas do not.
This pattern has practical research implications. If you want to understand what genuinely unites a diverse set of entities, meaning the underlying concept they all reference, navigating toward the shared high-in-degree neighbours in the graph often reveals the answer more efficiently than reading each entity's Wikipedia article and trying to abstract the common theme manually.
2D vs 3D View: When Each Is Useful
The Graph portal offers both a flat 2D network view and an immersive 3D rendering. The choice between them is partly aesthetic and partly functional.
The 2D view is better for close analytical work: reading cluster structures, comparing the relative proximity of specific entities, and taking mental notes about connection patterns. Because all nodes are visible on a plane, the eye can scan the entire rendered neighbourhood at once without rotating the view. This makes it easier to count connections, assess cluster density, and follow a specific path between two entities.
The 3D view is better for getting an intuitive sense of scale and structure. Switching to 3D when you are first exploring an unfamiliar part of the graph can reveal depth, layers of connectivity that the flat view compresses into a single plane. Some users find that the 3D view makes certain cluster boundaries clearer, because the z-axis separates entities that the 2D view stacks on top of each other.
The practical recommendation is to start in 2D for focused analysis and switch to 3D when you want a fresh perspective or when the 2D view feels overcrowded.
Practical Applications: Research, Education, Content Strategy
For academic researchers, the Graph portal offers a way to rapidly audit the conceptual neighbourhood of a topic they are writing about. Seeing which concepts are densely connected to your research subject, and which you had not previously considered, can reveal gaps in a literature review or suggest comparative cases worth examining.
For educators, the graph provides an immediate answer to the student question of why a topic matters: the density of its connections to other concepts they already know is a visual argument for its centrality. A history class exploring the Industrial Revolution can see its graph position and understand intuitively why it connects to economics, labour, urbanisation, colonialism, and science simultaneously.
For content strategists, the graph reveals which concepts serve as bridges between audiences. An entity with high in-degree that connects two otherwise separate clusters is a topic that speaks to both audiences, a natural meeting point for content that wants to reach across subject boundaries.
