What is a reading knowledge graph? A complete guide
A reading knowledge graph maps your library as a network: books are the points, and the connections between them (shared ideas, disagreements, links across subjects) are the lines. Each connection has a type, a strength and a reason, which turns a flat list of books into something you can think with.
The term is new enough that people use it loosely. This guide explains what it means, how it differs from a list or a catalogue, what kinds of connections belong in one, and what you can do with it, using Hikara's version as the worked example.
Updated By Hikara Team
What is a reading knowledge graph?
It's a graph in the computer-science sense: books are nodes and the relationships between them are edges. In a useful one, each edge has a type (what kind of relationship), a strength (how strong) and an explanation (why). Hikara's AI builds these from each book's title, authors and subjects plus your own ratings and notes.
You've used one without noticing: the panel beside a Google search result, listing a person's birthplace, books and relatives, is drawn from Google's Knowledge Graph. Knowledge graphs started as a general way to represent things and the relations between them.
Applied to reading, they let you ask questions a list can't answer: what in my library pushes back against this new book? Which two books I've read would a third one bring together? And when you half-remember an idea but not where you read it, following connections is often the quickest way back to the book.
How is it different from a reading list or a catalogue?
A reading list is a sequence: one book after another. A catalogue is a table: books sorted and filtered by fields like author, genre or tag. A knowledge graph is a network: any book can connect to any other, each connection says what kind it is, and you explore by following connections rather than scrolling.
A list answers "what have I read?" It can't answer which of your books reinforce each other, which actually disagree, which unlikely pair shares a thread you missed, or where the gaps in your reading are. Each of those is a question about relationships, and only a structure that stores relationships can answer it.
What kinds of connections go in it?
A few clearly different types work better than many. Hikara uses three: an echo (two books reinforcing the same idea), a challenge (two books that disagree or pull against each other) and a bridge (books from different fields that share an idea). Each connection gets one type and a strength from 0 to 100.
Why three? With two types you lose the difference between agreeing within a subject and carrying an idea across subjects. With many more, the categories start to blur and are hard to keep in mind. Three covers agreement, disagreement and transfer.
What can you do with one?
Three things a list can't do. See connections between books you've already read that you hadn't noticed. Find a next book that would bring together several you care about (Hikara's gap finder does this for two to five books). And spot one-sided reading: a cluster of books that only echo each other may mean you've read only one side of an argument.
Should you build one yourself or use a tool?
You can build one by hand, by linking book notes in an app like Obsidian, and it will be exactly yours, but every link is work and it grows slowly. A tool like Hikara finds and scores the connections for you as you add books. The trade-off is the usual one: control against effort.
FAQ
Does Hikara use a graph database?
No. Connections are stored as rows in PostgreSQL, each with its two books, a type, a strength and an explanation. For a personal library, a dedicated graph database would add complexity without adding anything you'd notice.
Is a reading knowledge graph the same as a personal knowledge graph?
Related, but not the same. A personal knowledge graph links your notes and ideas; a reading knowledge graph links books. They work well together: notes in Obsidian or Notion, and the books and how they relate in a reading tool.
How many books do I need?
A connection needs two books, so it starts working as soon as you have a few. Patterns, like themes that keep echoing or debates you've only read one side of, show up as the library grows.
Are AI-found connections reliable?
Treat them as suggestions with reasons attached, not facts. Hikara's AI is instructed to be honest about strength rather than generous, so weak links score low, and every connection comes with its explanation so you can judge it yourself.
Want to put this into practice?
Add your books to Hikara and see the echoes, challenges and bridges between them. The free plan needs no card.
Try Hikara free