similar pages uses machine learning and graph theory to help you find connections between ideas.
coming soon to roam depot.
for a given page in your graph, we calculate how similar the other pages are along two dimensions:
- how close together they are in the graph (using Dijkstra's shortest path)
- how similar they are according to machine learning (using Universal Sentence Encoder)
this is meant to help surface pages that are far apart, but similar:
after installing, click the new graph button next to the search input:
then select a page to get started:
everything is client side, your data never leaves roam. this plugin uses webworkers and indexeddb to make that possible.
Credit to Stephen Solka, creator of logseq-graph-analysis, which has served as a source of inspiration and guidance for representing a tft graph with graphology.