#embeddings

11M papers mapped by semantic similarity and time

reddit.com · ⭐️ 8/10 · 2026-07-01

8/10

A free interactive tool called The Global Research Space now visualizes 11 million academic papers using SPECTER2 embeddings and UMAP projection, with temporal slicing and daily updates. This helps researchers quickly grasp macroscopic trends across vast scientific literature, making it easier to stay current with daily paper floods. The map uses voronoi partitioning around high-density peaks at multiple depth levels, supports keyword and semantic queries, and includes an analytics layer for ranking institutions, authors, and topics.