What this shows
Unlike the text-based Formula Explorer (which groups by lemmas), this atlas uses Reciprocal Rank Fusion (RRF) across four verse-similarity algorithms (Jaccard, TF-IDF, Translation-pivot, Character-bigram) to identify the 200 most consistently-similar verse clusters. Each cluster carries its own geographic signature.
How to read it
- The sidebar lists all 200 clusters sorted by member count (default), by place count, or cross-lingual first.
- Click any cluster → the right panel shows the representative verse in a large poster, all 5 sample variants, the geographic map, and the top-place list.
- Circle colour indicates language: blue for Estonian, orange for Finnish, purple mixed, grey for unclassified.
- Language filter ET+FI cross isolates the 2 clusters where a single formulaic verse is attested in both language traditions in balanced numbers (suud ei kullata kuluta, siis ma ei maksnud maasta rohtu).
Caveats
- Only 5 sample variants per cluster are stored; deeper variant browsing lives in the main Reader.
- Places without a language tag render in grey and belong to uncertain or foreign collection points.
- The 200 clusters are the top RRF matches — there are thousands more weaker formulas not shown here.