Verse Path Finder

What this does

Finds the shortest chain of similar verses connecting two verses through the similarity network. Uses bidirectional breadth-first search — expanding from both endpoints simultaneously until the frontiers meet.

How similarity works

• Jaccard — shared exact wordforms between verse lines
• TF-IDF — shared lemmas (dictionary forms), weighted by rarity
• Translation — shared English translations, enabling cross-lingual Estonian-Finnish matches
• CharBigram — character bigram overlap, capturing surface-level textual similarity
• Sentence — sentence embedding similarity using all-MiniLM-L6-v2 on English translations

In "All (RRF)" mode, neighbors from all five algorithms are combined. In single-algorithm mode, only that algorithm's neighbors are followed.

Reading the path

• Each card is a verse; green endpoints are your selected verses
• The percentage on each arrow is the similarity score between adjacent verses
• Cross-corpus edges (ET/FI) are marked with a special badge
• Click any verse card to open it in the Verse Network explorer

Enriched search

When enabled, the path finder computes extra Jaccard (exact wordform) matches on-the-fly from all cached chunks, discovering connections that were below the pre-computed top-5 cutoff. This greatly improves connectivity for verses sharing vocabulary.

• Only discovers Jaccard (exact wordform) matches — not TF-IDF or Translation
• Only searches within already-loaded chunks (up to 30 cached)
• Enriched edges are marked with a Computed badge in hop details
• Disable for faster searches when pre-computed neighbors suffice

Tips

• Start with max hops = 3 or 4; increase if no path is found
• Enable "Enriched" mode to discover more connections through shared wordforms
• Using "Translation only" mode can find cross-corpus paths more easily
• Very common recurring verses act as "hubs" — paths often pass through them
• If two verses have no path within 7 hops with enrichment, they may be in disconnected parts of the similarity graph