Similar to Fedivertex: lemmy cross_instance

Ranked by cosine similarity over each dataset's title, summary and tags. Similarity is computed from descriptions, not from the data itself.