ELS research laboratory for the Masoretic Hebrew text: morphology-aware Deep Hunter, 3–6 ELS networks, rigorous control corpora, and reproducible experiments.
Deep Code Hunter · 3–6 nodesThe system freezes MT-derived candidates before ELS: surface forms, recurring phrases, OSHB lemma-family metadata, observed variants, and explicitly marked heuristic-root hypotheses. It then finds dense ELS networks and validates the winner against full control corpora.
A heuristic root is only a computed hypothesis and is always marked as such. With the full OSHB corpus, the interface shows the share of tokens with lemma/morph metadata.
No predefined wordsHunter mines a candidate lexicon from the selected Hebrew corpus before viewing ELS results and freezes it for all control runs.
For the full Tanakh, start with 60–100 candidates and 2–4 controls/model; increase the number of control runs under a pre-frozen plan for confirmation.
Automated experimentAll words in the selected lexicon are tested simultaneously; pairs are not predefined.
Increase the number of control runs for a serious result. All parameters and the seed are saved with the corpus SHA-256.
Discovery first computes ELS and clusters locally. AI is used only after statistical ranking.
Biblical Hebrew dictionaryHebrew–Russian and Hebrew–English lexicon linked to OSHB lemmas: Strong’s, abridged BDB, morphology, and every occurrence in the loaded corpus. Proper names show their meaning, who is related to whom, and a compact family tree. Works locally without AI.
Search by Hebrew with or without vowels (inflected forms are recognised), transliteration, English or Russian meaning, a proper name (Adam, Адам), or Strong’s number. Double-click a Hebrew word in any result to look it up.
⚠ An ELS match alone does not prove intentional encoding. Research Grade measures how unusual the discovered “winner” is relative to full control runs.