# UECF Hebrew – Greek – Latin – Telugu – English Parallel Bible A five-column parallel Bible reader for uecf.net, showing all 66 books of the Protestant canon in Hebrew (transliterated), Greek (transliterated), Latin (Vulgate), Telugu (Unicode), and English (NKJV), aligned verse by verse. ## Sources | Column | Source app | Script | |---------|----------------------|-------------------------------------------| | Hebrew | `romanhebrew` | Romanized (Latin-script) transliteration | | Greek | `romangreek` | Romanized (Latin-script) transliteration, Old Testament portion rebuilt for this project (see below) | | Latin | `latin` | Native Latin (Clementine Vulgate) | | Telugu | `telugubible-latin` | Native Telugu Unicode (despite the source folder's name, its actual text is Telugu script, not romanized -- verified by direct inspection; "latin" in that folder's name refers to its Vulgate/Catholic-canon book list, not its script) | | English | `nkjv` | New King James Version | None of the five source app directories were modified. Every verse of Hebrew, Latin, Telugu, and English text here is byte-identical to its source app's own verse. The Greek column's New Testament text is likewise byte-identical to `romangreek`'s existing NT data (SBLGNT-derived, already accurate); its Old Testament text was rebuilt from a different, more reliable source than `romangreek`'s original OCR pipeline -- see "Why the Greek Old Testament was rebuilt" below. Hebrew and Greek are shown in their existing Romanized form, as supplied, and were **not** auto-converted to native Hebrew/Greek Unicode script. ## Why the Greek Old Testament was rebuilt While building this app, cross-checking `romangreek`'s Old Testament verse counts against NKJV's revealed that `romangreek`'s Septuagint text (itself OCR'd from a 147-year-old scanned 1879 edition with no digital text layer) had severely broken verse boundaries throughout -- e.g. its "Genesis chapter 1" contained 44 mis-cut fragments instead of 31 real verses, with individual verses ending mid-word. This is a pre-existing defect in that OCR pipeline (documented at length, and honestly, in `greek-old-testament/site/reports/known-limitations.txt`), not something introduced by this project. The New Testament (SBLGNT-derived, extracted from a PDF with genuine embedded text) was unaffected and is used as-is. Rather than build this 5-language site on top of known-broken verse alignment for 39 of its 66 books, the Old Testament Greek column was rebuilt from **LXX-Rahlfs-1935** (Alfred Rahlfs' 1935 critical edition of the Septuagint, digitized by Eliran Wong / the Open Scriptures Septuagint Project, ultimately derived from the CATSS/CCAT academic Greek corpus): . That dataset provides real, correctly-inflected Greek surface text (not just dictionary lemmas) with reliable, independently-sourced verse boundaries. **License and attribution.** This dataset is licensed **CC BY-NC-SA 4.0** (Attribution-NonCommercial-ShareAlike), copyright Eliran Wong, based on work from the CATSS/CCAT Septuagint project and the Open Scriptures Septuagint Project. UECF.NET's use here is as a non-commercial church ministry site, consistent with this license. Required attribution: *"Greek Old Testament text derived from LXX-Rahlfs-1935 (Alfred Rahlfs' 1935 Septuaginta), © Eliran Wong / Open Scriptures Septuagint Project, based on CATSS/CCAT data, licensed CC BY-NC-SA 4.0."* This attribution must stay with any redistribution of this specific Greek Old Testament text or its transliteration. **Method** (`tools/rebuild_greek_ot.py`): the dataset supplies (a) a flat, sequential word-by-word list of the entire LXX corpus in accented Unicode Greek, and (b) a parallel list of verse-start markers (`Book chapter:verse` -> starting word index) aligned to that word list. Each verse's text is simply the words between its own start marker and the next verse's start marker -- no text was rewritten, corrected, or invented; every word is exactly as the source dataset has it, only correctly grouped into verses. The result: Genesis 1 now has 31 correctly-bounded verses (compare `romangreek`'s 44 broken fragments before this fix); across all 39 Old Testament books, 7 books now match NKJV's chapter/verse structure exactly and the rest are close, with genuine, well-documented ancient Hebrew-vs-Greek versification differences (not defects -- see "Known versification differences" below) accounting for most of what remains. The romanized ("English script") form of this rebuilt Greek text was then produced with the same `translit.py` engine already used for `romangreek`'s New Testament, unchanged. **Text-tradition choices.** Rahlfs prints more than one Greek textual tradition for a few books; exactly one was chosen per book so this site has a single coherent verse per reference. These choices are recorded, with the reasoning, in `tools/rebuild_greek_ot.py`'s `BOOK_MAP` and echoed in `reports/alignment_report.txt`: - **Joshua, Judges:** the "B" (Codex Vaticanus) textual tradition (Rahlfs prints both "A" and "B" texts for these two books; "B" is the more complete of the two in this corpus). - **Daniel:** Theodotion's Greek translation (`DanTh`), the version traditionally printed as "the" Septuagint Daniel in editions such as Brenton's -- the same edition family as this site's original OT PDF source -- rather than the Old Greek (`Dan`). - **Ezra / Nehemiah:** Rahlfs' "2 Esdras" (23 chapters) combines both books; this site splits it at the conventional boundary (chapters 1-10 = Ezra, 11-23 = Nehemiah, renumbered 1-13) to match the other four languages' book division. - This rebuild does **not** touch `romangreek`'s live site data. The rebuilt 66-book Greek data set (both native Unicode and transliterated forms) is available as a ready-to-deploy fix for `romangreek`/ `greek-bible` if UECF.NET wants to apply it there too -- ask for it separately; it was left out of this delivery to keep this folder to just what this app needs (it added ~25 MB of files this app never reads). That would be a separate, not-yet-applied deployment. ## Scope: 66-book Protestant canon `romanhebrew`, `romangreek`, and `nkjv` each contain exactly the 66-book Protestant canon. `latin` and `latin`'s Telugu Catholic-canon counterpart contain 73 books (the same 66, plus 7 deuterocanonical/apocryphal books: Baruch, 1-2 Maccabees, Judith, Sirach, Tobit, Wisdom). This site covers the 66-book canon that all five sources share; the 7 extra books are out of scope for this 5-way alignment, since three of the five columns would be entirely blank for them. This matches the spec's own "at minimum Genesis-Malachi, Matthew-Revelation" scope. ## Known versification differences (documented, not silently aligned) Ancient translations don't always divide chapters and verses identically. Where a genuine numbering difference exists between two Bible traditions, this site does **not** force verses into artificial alignment -- each chapter's verse list is the union of every verse number any of the five sources uses for that chapter, and a language with no text at a given verse number shows "—" rather than being renumbered. The most notable cases (all inherited from the underlying texts, not artifacts of this build): - **Psalms:** the Greek (LXX) column keeps its own traditional numbering, which runs one behind the Hebrew/English numbering for most of the book (LXX merges Hebrew Psalms 9-10 and 114-115, and splits Hebrew Psalms 116 and 147 each into two) -- e.g. what this site's Hebrew, Latin, Telugu, and English columns call Psalm 23 lines up with the Greek column's own Psalm 22. - **Malachi:** the Greek/Latin tradition has only 3 chapters; English chapter 4 corresponds to the end of Greek/Latin chapter 3. Malachi 4:2 is one of this project's required test passages and is a direct example of this: the Greek column shows "—" at that specific chapter:verse key (the text exists, just under chapter 3 in that column) while the other four languages show real text. - **Esther, Daniel:** the Greek text includes additional material beyond the Hebrew/English text in these two books. - Full per-book, per-chapter detail is in `reports/alignment_report.txt` and `reports/alignment_report.json`. ## Data structure Each book file (`data/books/.json`): ```json { "id": "genesis", "number": 1, "testament": "OT", "nameEn": "Genesis", "nameHe": "Bereshit", "nameGr": "Genesis", "nameLa": "Genesis", "nameTe": "ఆదికాండము", "shortName": "Genesis", "chapterCount": 50, "chapters": [ { "n": 1, "verses": [ { "n": 1, "he": "...", "gr": "...", "la": "...", "te": "...", "en": "..." } ]} ] } ``` `data/meta.json` lists all 66 books (id, number, testament, all five name fields, chapterCount, per-chapter verseCounts). `data/search.json` is a flat array of `[bookIndex, chapter, verse, he, gr, la, te, en]` rows. ## Regenerating the data 1. `tools/rebuild_greek_ot.py` -- rebuilds the Greek Old Testament from LXX-Rahlfs-1935 (requires that dataset checked out locally; only needs re-running if the Greek OT source changes). 2. `tools/merge.py` -- reads all five source apps' `data/meta.json` + `data/books/*.json` (edit the `SOURCES` paths at the top of the script to point at wherever those five apps live on the machine you're running it from) and writes this app's `data/`, plus `reports/alignment_report.{json,txt}`. Neither script modifies any source app's own data. ## Search The search box matches word/phrase text across all five languages by default, or a single language via the "Search:" dropdown on the search page. A recognized `Book chapter:verse` reference (e.g. "John 3:16") jumps straight to that verse; a recognized book name jumps to its first chapter. Results link directly to the matching verse. ## Responsive layout - **Desktop / laptop (>= 1180px):** all five columns shown at once, verse-aligned in a single row, with a sticky Hebrew/Greek/Latin/ Telugu/English column header. - **Tablet (769-1180px):** the same five-column grid, but the verse list scrolls horizontally with the verse number pinned in place, so no column's text is shrunk to stay readable. - **Phone (<= 768px):** each verse becomes a stacked card with all five languages labeled and listed vertically -- chosen over horizontal scrolling or shrinking text, per the spec's "prioritize readability" guidance. ## Deploying Upload every file and folder in **this directory** (`index.html`, `.htaccess`, `css/`, `js/`, `data/`, `reports/`, `tools/`, and the 66 per-book `.html` redirect stubs) directly into `/hebrew-greek-latin-telugu-english/` on the web server -- do not upload this folder itself as an extra nested level, and do not upload only a subfolder of it. `tools/` and `reports/` are not required for the site to function but are harmless to leave in place (matching the existing `hebrew-english-parallel-bible` app's own deployed layout, which keeps its `tools/` and `reports/` alongside `index.html`). ## Validation `reports/alignment_report.json` / `.txt` are generated automatically by `tools/merge.py` on every run -- they report per-book/chapter verse counts, every missing-verse and duplicate-verse case by language, every whole-book chapter-count mismatch, and pass/fail on the 13 required test passages (Genesis 1 & 3, Psalm 23, Isaiah 53, Malachi 4, Matthew 1 & 5, John 1, 3 & 14, Romans 8, 1 Corinthians 13, Revelation 21). Current totals: 66 books, 1,204 chapters, 32,697 verse-rows; 29,473 (90.1%) have all five languages present, 3,224 have at least one language missing at that exact chapter:verse (almost entirely explained by the versification differences above); zero duplicate verses in any language. ## Limitations - Hebrew and Greek are transliterations (Latin-script phonetic representations), not the original scripts -- by design, per the project spec. - The Greek Old Testament's text-tradition choices (Joshua/Judges "B" text, Theodotion Daniel) are documented above; a different, equally valid scholarly choice was possible for each. - `latin` and `telugubible-latin`'s 7 deuterocanonical/apocryphal books are out of scope for this 5-way view (see "Scope" above). - `data/search.json` (the full-text search index, all 5 languages x 32,697 verse-rows) is about 25 MB. It is fetched lazily, only the first time a reader actually searches -- normal chapter reading never loads it -- but that first search will be slow on a poor connection. Splitting it into per-testament or per-language shards would reduce this; not done here to keep this build a direct extension of the existing UECF.NET `search.json` pattern used by every other Bible app on the site. - This is a reading/reference tool, not a critical edition -- readers needing guaranteed letter-perfect text for citation should cross-check against a printed edition, as `romangreek`'s and `latin`'s own documentation already advises for their source material.