I created a map with the keyword "information".
A few odd categories returned:
- Scientific publishing, Metacognition, Dogs
- Twitter, Nutrition, Socioeconomic aspects of health
- Graphs, Quality of life
When I've thought about systems like this one, I always get stuck on that problem. If you automatically create every link between nodes, you end up with terrible signal-to-noise for the user; even 50 links (I'm guessing at a threshhold) seems too many to read through in order to find the one you want, and too many for the user to conceptualize. Imagine how many links there are to nodes like evolution or relativity.
The solutions I can think of:
* Categorize links, so the user can narrow their search
* Rate links by value? Relevance? How do you do this without a human editor?
* Efficient user-created filters for links
* Directional links: e.g., Many nodes may link to relativity, but relativity links to only a few
* Expert human editor, who can identify valuable links, but that wouldn't seem to scale
* Crowd-sourced links, but I would expect quality and signal-to-noise to be too low for serious research
IME, for serious research, nothing automated approaches the value of a professional reference librarian.
I searched "computer", and it came up with some strange results. Poking around a bit, it says it's based on "PLOS", which after jumping through that link and poking around some more it seems to be mostly for biology/medical related fields of research.
We have also currently deployed a simpler algorithm due to performance issues on our server.