Study 2 / Study 4 substrate

The equivalence index

For any Hebrew word, what Greek renders it; for any Greek word, what Hebrew stands behind it — counted, two-way, and filterable, over the CATSS Hebrew–Greek alignment.

In print its ancestors are Hatch–Redpath's Concordance to the Septuagint (1897, Greek → Hebrew) and Muraoka's Hebrew/Aramaic Index to the Septuagint (2010, the reverse, still in copyright — consulted, never embedded). This is a two-way version of the same idea over an open text: 4,044 Hebrew lemmas and 3,494 Greek lemmas, drawn from four books on disk — Esther, 1 Kings, Job and Daniel (Old Greek) — not yet the whole Septuagint.

Every rendering here comes from the CATSS parallel alignment (Center for Computer Analysis of Texts, University of Pennsylvania), used as an index only: no CATSS line is reproduced, and the counts and reference lists are our own facts about two public-domain texts. Full method and sourcing: colophon.

How reliable this is

The Hebrew side is lemmatised by Open Scriptures' Hebrew Bible (CC BY 4.0). The Greek side is the edition's own: every one of the four books' distinct Greek surface forms was read and lemmatised by a language model, in context, and checked on a blind 500-form sample against CCAT's own lxxmorph morphological analysis (that comparison text is never published, per the CATSS declaration) — 100% adjudicated agreement once dictionary-citation-convention differences (a compound cited fused vs. decomposed into stem and preverb, a deponent verb cited by its active vs. middle form) are set aside from genuine errors. Six of ten literature-attested pairs have been spot-checked against Hatch–Redpath/Muraoka and land where expected — see the table below.

Sub-pairing inside a multi-word CATSS line (splitting "in the days of" into its three Hebrew and Greek parts, say) is mechanical for conjunctions, articles, prepositions and the object marker; a full model pass over the remaining "bundle" lines was sampled at 1,000 of 9,065 lines and found reliable, but has not yet been run at full scale or merged into this index — so some renderings below are coarser (a whole multi-word span credited to a whole multi-word span) than the finest possible grain.

Six spot-checked pairs

The sixty most frequent pairs

By total occurrences across the four books. Browse any Hebrew or Greek word directly, or use search — every one of the 4,044 Hebrew and 3,494 Greek pages is indexed.

Data: every unit as JSON includes equiv-he.json, equiv-gr.json and equiv-stats.json.