Source: Supreme Court of the United States - Wikipedia

Caveat: the maps below are likely wrong. Please don't use them as examples of good concept maps. They're only shared publicly to document the tool's evolution.

Second Try

Supreme Court of the United States - A concept map by LLMapper

Struggling with camelCase labels, an artifact of the RDF step.

Fifth Try

Supreme Court of the United States - A concept map by LLMapper

What’s worse than rendering camelCase? Doing so in a scattered map.

Sixth Try

Supreme Court of the United States - A concept map by LLMapper

Better, but there are still some issues “President Nominates” shouldn’t be a node label. (The edge label “nominates” is correct.)

Comments

One of LLMapper’s current limitations is that it focuses on the subject’s components without delving into why it matters in a broader context. For example, a map about the U.S. Supreme Court would benefit from explaining the role of the judicial branch in U.S. governance. You don’t get that in any of these results. That’s surely an effect of the first prompt, which merely instructs the LLM to summarize the Wikipedia page and list its main concepts.

Changes

  • Instructed DOT step to convert phrases from camelCase and other non-standard phrasings to normal English phrases (to mixed results)
  • The DOT step sometimes opens with the Markdown comment block. I’ve tried several approaches to get it to ignore this, but it’s still sometimes being rendered. I might have to return to removing it with regex via the shell.