651 instances of 334 editions across five languages, with verified canonical metadata.
1 Corinthians 13:12
Βλέπομεν γὰρ ἄρτι δι᾽ ἐσόπτρου ἐν αἰνίγματι
we-see · for · now · through · mirror · in · riddle
"for we see now through a mirror obscurely"
"For now we see in a mirror, darkly"
"For now we see through a glass, darkly"
"For we see now through a dim window obscurely"
"For now we see in a mirror, dimly"
Prior corpora typically pick one canonical text per translation family. Each verified edition is retained — the 1611 King James and the 1769 Cambridge revision are different texts, and a corpus that collapses them loses the diachronic signal that makes within-family comparison interesting.
26% of works carry more than one edition, but those works account for 51% of the editions — roughly half the corpus would disappear under one-canonical-per-work deduplication.
Public-domain and openly-licensed editions ship as plain text via GitHub and Hugging Face. Copyrighted editions are available on request for non-commercial research.
For every edition — including the copyrighted ones — pre-computed verse- and chapter-level embeddings and full pairwise similarity matrices (lexical and semantic) are published, so downstream work on the corpus's structure doesn't require the source text.
Browse what is available, read individual translations, or compare passages side by side.
Download corpus texts, embeddings, and pairwise similarities.
137 editions in the public release; copyrighted on request.
Verse- and chapter-level, Qwen3-Embedding 0.6 B and 8 B.
Lexical and semantic, for all within-language edition pairs.