I made this knowledge extraction pipeline to produce a graph of concepts from scientific papers, surfacing connections across fields that might otherwise go unnoticed. I tested this on a pilot set of papers from the neurobiology of vocal learning and sensorimotor learning (my doctoral/post-doctoral research area), and right away found some insights that seem valuable.

Interestingly, the same pipeline extracts useful entities and relationships from lore books about Final Fantasy, showing that the method works generally across domains. It'd be great to hear what anyone thinks.