Obsidian is a popular note-taking app that helps users connect their ideas using the [[wiki links]] syntax. However, the default graph view offered in the app as well as the third-party graph view plugins are not really useful. They do visualize connections between the pages but lack the structural insights that can be obtained using network analysis.

To rectify this problem, we created a graph view plugin for Obsidian which uses graph theory and network science algorithms to rank the ideas and pages by their relative influence, using the betweenness centrality measure. Additionally, an advanced force-layout algorithm creates a 3D graph view layout that pushes the communities detected using the modularity algorithm apart from each other, making it easier to see the clusters of ideas that belong together. As a result, it is possible to calculate distances between these clusters and reveal the structural gaps in the knowledge graph. These structural gaps represent the blind holes in the discourse and can be used to generate new ideas (for example, using the built-in GPT-4 AI).

The Problem with Obsidian Graph View

One of the main advantages of Obsidian that sets it apart from other note-taking apps is its ability to connect different notes. It’s possible to highlight a concept using the [[wiki links]] syntax, which will create a file / page with the same name. This page will list all the “linked” (proactively [[highlighted]]) and “unlinked” (plain text mentions) references to itself. As a result, it becomes much easier to find connections between different ideas, so that they’re not lost.

The native graph view of Obsidian visualizes connections between the pages, but it doesn’t take into account any structural insights about the graph. You cannot really see which nodes / pages are more important, what are the groups they form, and what are the gaps between them. It may be possible to retrieve all those insights just by looking at the graph, but it is quite hard because the graph itself doesn’t have any data or rankings:

Other alternatives, such as the Juggle graph view community plugin for Obsidian, are a little better, but still lack the structural insight and don’t visualize the connections between the pages that appear in the same context.

One of the reasons is that the default Obsidian graph view and all the plugins only show connections between the main page and the the [[wiki links]] mentioned on that page. However, it doesn’t show the connections of those [[wiki links]] to each other. For instance, if you have a page called “Money” and you write: “As [[inflation]] increases, [[interest rates]] rise”. You will only have a connection from [[Money]] to [[inflation]] and from [[Money]] to [[interest rates]], but you won’t see the connection between [[inflation and [[interest rates]], which is actually the most relevant and interesting relation. As a result, the graph often has a star structure, so it’s impossible to provide any structural insights because all the nodes that connect to the main “central” one look the same.

The only useful use case for the Obsidian graph view plugin is to see the whole vault at once as it provides some sort of aesthetic pleasure of watching your knowledge grow. You can also see the main pages that have the most connections, but that is as useful as the graph will get.

Some people find it useful to introduce colors based on the pages’ properties, but it’s a lot of manual work and it still doesn’t even attempt to tap into the vast repository of insights that text network graphs can offer.

InfraNodus 3D Graph View Plugin

To address the problems indicated above, we decided to create our own version of Obsidian graph view plugin. It uses InfraNodus text network algorithm to build a graph of the connections between the [[wiki links]] and the concepts identified in a text and to apply advanced graph theory algorithms to rank nodes and clusters. As a result, you get a much more precise picture of your knowledge and can also see how all the ideas are connected to one another.

One big difference from the other graph view plugins offered for Obsidian is that InfraNodus will show connections between the [[wiki links]] mentioned in the same paragraph. So if we use the example above, if you have a page called “Money” and you write: “As [[inflation]] increases, [[interest rates]] rise”, InfraNodus will not only make a connection from [[Money]] to [[inflation]] and from [[Money]] to [[interest rates]], but also between [[inflation and [[interest rates]] (which was the whole point of writing this statement in the first place).

Additionally, InfraNodus will by default also visualize the concepts used in a text as the nodes and their co-occurrences as the connections (this can be switched off in the settings). Therefore, the users get a much richer graph that has more relational information and, as a result, it’s possible to retrieve more structural insights about the underlying text network. The more influential concepts are shown bigger on the graph and if the concepts (or wiki links) tend to appear often in the same statements or paragraphs they will have the same color. This provides a rich structural overview of ideas that help users see which ideas are most relevant and what clusters they form so that it’s easier to identify the topical clusters. As a result, it is also possible to see which topical clusters have gaps between them and generate new ideas by asking the questions that would connect the ideas from those clusters in a new way:

InfraNodus Obsidian graph view plugin also has an AI module based on GPT-4 model. It’s possible to select a certain aspect of discourse using the graph (or the topics or gaps) and to then generate a summary or a review question that would link these concepts together. As the plugin does not store any information on the server, it offers a highly private way to generate ideas from your private notes without sending the whole vault to an AI model.

…

If you’re interested in trying the new InfraNodus Obsidian graph plugin, please, find the download instructions on InfraNodus.Com and the installation instructions on our support portal.

…