Skip to content
On this page

Knowledge Graph

A visual map of concepts in your research area. See clusters, gaps, and connections at a glance.


What It Is

The Knowledge Graph is an interactive 2D force-directed graph where:

  • Nodes = key concepts (e.g., "transformer", "attention", "time series")
  • Edges = relationships (co-occurrence in your papers)
  • Size = frequency of the concept

Think of it as a social network for ideas.


When to Use It

  • After collecting 5+ papers with insights
  • To identify sub-fields within your topic
  • To find isolated clusters (unexplored intersections)
  • To decide where your research could make a novel contribution

How to Generate

  1. Go to Insights
  2. Ensure you have several papers with insights
  3. Click Generate Knowledge Graph

The graph loads in a few seconds. You can pan, zoom, and drag nodes.


Reading the Graph

  • Tightly connected cluster: Well-studied area; hard to contribute something new
  • Sparse nodes: Concepts that appear alone—maybe a niche worth exploring
  • Bridge nodes: Concepts that connect clusters are good candidates for interdisciplinary work

Hover a node to see which papers mention it.


Customization

Currently, the graph is auto-generated from your paper insights. You cannot manually add/remove nodes (future versions may allow curation).

Switch projects to see different graphs.


Example

Suppose you search for "deep learning for weather forecasting". The graph might show clusters around:

  • RNNs and LSTMs (older methods)
  • Transformers (newer)
  • Numerical weather prediction (physics-based)

If you notice a small bridge between transformers and numerical methods, that could be your research gap.


Limitations

  • Depends on quality of AI-extracted concepts
  • Works best with at least 10 papers
  • No filtering yet (coming soon)

Next

Put your graph to work by staying organized with Kanban and Grants.

Questions? Open an issue or join our Discord.