README.md

---
license:
  - cc0-1.0
  - cc-by-4.0
pretty_name: "Hebrew Lexical Reference Indices — BDB roots, Hebrew WordNet synsets, T'OMIM parallelism, LXX bridge"
language:
  - hbo
  - grc
tags:
  - biblical-hebrew
  - biblical-greek
  - lexicography
  - reference-data
configs:
  - config_name: bdb_roots
    data_files: "bdb_roots.tsv"
  - config_name: hwn_synsets
    data_files: "hwn_synsets.tsv"
  - config_name: tomim_parallelism
    data_files: "tomim_parallelism.tsv"
  - config_name: lxx_bridge
    data_files: "lxx_bridge.tsv"
---

# Hebrew Lexical Reference Indices

Four structured, Strong's-linked transcriptions of external Hebrew (and one Hebrew↔Greek) lexical
reference sources. **These are not our own synonymy judgments** — each config faithfully represents
what an established outside source, or an actual historical translation record, already asserts (an
etymological dictionary's own root groupings, a WordNet's own synset membership, five named scholars'
own verified structural analysis, the Septuagint's own attested rendering choices). We didn't decide
these pairs are related; the cited sources did. That distinction matters: read each config's
description below before treating a pair as "verified synonyms" — none of these four, on their own,
make that specific claim.

## `bdb_roots` — Brown-Driver-Briggs etymological root groups

Hebrew Strong's numbers grouped by the root Brown-Driver-Briggs (1906, Public Domain) assigns them to,
digitized from [OpenScriptures/HebrewLexicon](https://github.com/openscriptures/HebrewLexicon)'s
`LexicalIndex.xml` (CC BY 4.0 digitization, pinned commit `21c9add1`). 1,432 root-groups, 4,616 distinct
Strong's. This is BDB's own etymological classification, transcribed — not a claim that every pair in a
root-group is synonymous (etymological kinship and semantic proximity are related but different things).

| column | meaning |
|---|---|
| `root_id` | BDB's internal root identifier |
| `strong` | Hebrew Strong's number (`H####`) |
| `xlit` / `gloss` / `pos` | transliteration, gloss, part of speech, from BDB |

## `hwn_synsets` — Hebrew WordNet synset membership

Hebrew Strong's numbers sharing a synset in Hebrew WordNet (Ordan & Wintner, University of Haifa, 2007
— permissively licensed, "permission to use, copy, modify and distribute... for any purpose and
without fee or royalty"), bare-consonant-matched from HWN's modern-Hebrew lemmas to biblical Strong's
numbers via `lexeme-spine.db`. HWN is modern Hebrew, not biblical — matches reflect vocabulary that
survives into (or from) contemporary usage; biblical-only senses may not be covered.

| column | meaning |
|---|---|
| `strong_a` / `strong_b` | Hebrew Strong's numbers sharing an HWN synset |
| `hwn_synset_id` | the Hebrew WordNet synset identifier |

## `tomim_parallelism` — expert-verified poetic-parallelism pairs (T'OMIM-confirmed tier only)

Words that five named Hebrew-poetry scholars (Berlin 2008, Fokkelman 2001, Kugel 1981, Watson 1994,
Tsumura 2023), via the T'OMIM corpus (Smiley 2026, CC BY 4.0, DOI
[10.5281/zenodo.19135731](https://doi.org/10.5281/zenodo.19135731)), verified occupy matching slots
across parallel half-verses in Biblical Hebrew poetry — a real, human-vetted structural correspondence.
1,676 pairs. This is the **confirmed tier only** — a separate, unpublished "detected" tier (our own
BHSA-based extraction, not expert-verified) exists internally and is deliberately excluded here.

| column | meaning |
|---|---|
| `strong_a` / `strong_b` | Hebrew Strong's numbers in matching parallel slots |
| `relation` | `likely_synonym` / `likely_antonym` / `unclassified` — our own downstream classification of the verified pairing; `unclassified` means the structural correspondence IS verified, only the synonym-vs-antonym call is unresolved |
| `sample_ref` | one verse reference where this pairing occurs |

## `lxx_bridge` — OT Hebrew → NT Greek correspondence via the Septuagint

Which Greek Strong's number the Septuagint (LXX) translators actually rendered each Hebrew Strong's
number as, with attestation counts — a real historical translation record, not a similarity judgment.
Computed from MACULA's own Hebrew/Greek Strong's tagging (CC BY 4.0) aligned against the ancient
(Public Domain) LXX text. 3,301 rows, 1,862 distinct Hebrew Strong's, 1,571 distinct Greek Strong's,
noise-filtered to count >= 3. Hebrew and Greek Strong's don't cross-link in raw MACULA at all — this is
the only place that correspondence exists in structured form.

| column | meaning |
|---|---|
| `hebrew_strong` / `greek_strong` | the Hebrew word and the Greek word the LXX rendered it as |
| `count` | how many times this specific Hebrew→Greek rendering is attested across the LXX |

## License

`bdb_roots`: CC BY 4.0 (OpenScriptures digitization of Public Domain BDB). `hwn_synsets`: permissive
(Ordan & Wintner, U. Haifa — see above). `tomim_parallelism`: CC BY 4.0 (T'OMIM). `lxx_bridge`: CC BY 4.0
(MACULA Hebrew/Greek tagging; underlying LXX text itself is Public Domain). None of these four derive
from or reproduce UBS MARBLE / Louw-Nida / SDBH.

## Provenance

Built by `shoresh/macula/build_bdb_roots.py`, `build_hwn_benchmark.py`, `build_parallelism_pairs.py`,
and `build_prior_pack.py` (the `lxx_bridge` source) in
[bcv-commons/bcv-query](https://github.com/bcv-commons/bcv-query). See also
[bcv-commons/semantic-neighbors](https://huggingface.co/datasets/bcv-commons/semantic-neighbors) for
the separate, actually-verified synonym-pairs dataset these sources feed into as internal signals
(alongside others) — that dataset makes synonymy claims; this one does not.