AI empowered Zettelkasten with NER and Graph LLM

Knowledge management
1

Last time, we tweaked Obsidian to work with links and make it closer to Personal Knowledge Graphs with plugins.

Today, I want to explain how a classical Knowledge Graph, as an extra layer, could empower Knowledge Management with AI tools.

Idea is simple you have your idea zettelkasten + knowledgegraph generated by #ai entityrecognition (NER) and graphconstruction llm slm

As a result, you have a deeper and better connection and source for #link prediction and reasoning

We do it already for you #ai #agent conversation at mykin.ai

2 Likes
2

I made an article with more details AI-empowered Personal Knowledge Graphs in Obsidian. | by Volodymyr Pavlyshyn | Mar, 2024 | Medium

3

I really wish the community would get behind making links more like full-fledged edges in a graph sense. the link should have metadata.

1 Like
4

@vop Is there a followup to this work? I’m very interested in something like this. Well, not as far, but I think it would already be helpful to:

  1. Run an NER model to ensure entities are links for all notes (or some notes, as defined by their yaml metadata)
  2. Feed the extracted NERs to develop an ontological model about how the enties might be connected, and import that into Obsidian too. To @nickknyc 's point, it sucks that we can’t name the edges, but still having the ontological and logical alignment would be quite powerful, no?