All 27.9M embeddings (20.3M LC records + 7.6M Wikipedia abstracts) quantized into ~4,000 topic cells (spherical k-means), laid out so similar topics sit together (t-SNE of cell centroids). Circle area = number of records in the cell. Scroll to zoom, drag to pan, hover any cell for its contents and a real exemplar from each corpus. Click any cell to list up to 1,000 real titles from each corpus below, linked out to the LC catalog and Wikipedia.