Added on 3 October 2026 (docs/research/new-texts-scope.md, Part A, A.3): the reader's first continuous Persian verse. A first selection, chosen to pair with Ibn Sīnā's Risālat al-Ṭayr (/read/tayr) and the mirror path (docs/research/heart-sight.md).
| Work / author / dates | Manṭiq al-ṭayr, Farīd al-Dīn ʿAṭṭār of Nīshāpūr (d. c. 618/1221); mathnawī, about 4,500 couplets |
| Part added | three passages, 282 couplets (Tassy's numbering): the hoopoe's call vv. 658–724; the seven valleys vv. 3202–3228, 3313–3334, 3456–3481, 3558–3580, 3673–3680, 3779–3791, 3920–3936; the thirty birds and the Sīmurgh vv. 4104–4153, 4201–4229 |
| Reader route | /read/mantiq-tayr · about /read/mantiq-tayr/about · sections #hoopoe, #valleys, #simurgh · couplets #v-<verse> |
| Catalogue kind | poetry (with titleLang: "fa") |
| Status | working edition · transcription AI-read from the page images, human proofreading pending · reviewed by: — |
1. The licence question, resolved first
| Candidate | Finding | Decision |
|---|---|---|
| Ganjoor (ganjoor.net/attar/manteghotteyr) | No licence is stated for the texts. The FAQ calls the data "open-text and free" (bāz-matn va rāygān) without naming a licence; the desktop software is MIT on GitHub (ganjoor/desktop). The book's "paper sources" in Ganjoor's own API (Shafīʿī-Kadkanī; Gowharīn 1342/1963; Ḥamīd Ḥamīd) are all flagged isTextOriginalSource: false, so the base edition of the digital text is not named [VERIFIED 3 Oct 2026, API page?url=/attar/manteghotteyr]. | Unclear: not shown. Used only as a reading aid while transcribing; kept in data/raw/persian/attar/ganjoor/ (gitignored). |
OpenITI 0618FaridDinCattar.MantiqTayr.PDL00037-per1 | Listed in the metadata, not on GitHub (new-texts-scope A.3). Derived from the Persian Digital Library / Ganjoor. | Not used (same unclear base). |
| Garcin de Tassy 1857, Mantic Uttaïr … publié en persan (Paris: Imprimerie impériale) | Public domain (Tassy d. 1878). Not on archive.org: the item manticuttaroule00tassgoog, catalogued as 1857 Persian, is the 1863 French translation (its OCR has no Persian; checked). Found on BnF Gallica ark:/12148/bpt6k1183355j (184 pp., IIIF manifest saved) and Google Books cgdeAAAAcAAJ [VERIFIED]. Page images only; no OCR. | Base text. Transcribed for this selection from the Gallica images. |
Indian lithographs (DLI in.ernet.dli.2015.484948, 1883; …404895, 1891) | Public domain; images only. | Not used yet; a further witness for collation. |
So the public layer is our transcription of a public-domain print. Nothing from Ganjoor is published: where Tassy and Ganjoor differ, Tassy is followed (over half the couplets differ in at least a word or a spelling, many in whole words: v. 690 bar dirakhtī bas buland ārām-i ū against Ganjoor's dar ḥarīm-i ʿizzat ast; v. 3219 az halāk … nūr-i pāk against az ṣifāt … nūr-i dhāt; v. 4207 chun nigah kardand ān sīmurgh būd / bī shak ān sīmurgh ān sī murgh būd against Ganjoor's order ān sī murgh … īn sī murgh ān sīmurgh). Ganjoor's editorial vowel marks and punctuation are not used: Tassy prints almost none.
2. Transcription
- Source file:
scripts/persian/tassy-1857.tsv(verse, hemistich 1, hemistich 2, printed page, flag), with Tassy's chapter headings (#H, e.g. al-maqāla al-thāniya sukhan-i hudhud bā murghān dar ṭalab-i sīmurgh), section starts (#S) and gaps (#G). - Method: each page image (Gallica IIIF, 1400 px) read in three bands by AI (Claude), Ganjoor's couplet beside it as an aid; every word follows the image. Verse numbers were checked against Tassy's marginal numerals (every fifth verse) on every page. Canvas = printed page + 18 (pp. 26–28, 127–167 checked).
- Normalisation: word spacing (Tassy's joined forms yakdigar-rā, bi-rāh kept as printed where they are one word in his line), ی/ک, no vowel marks; hamza on -ye written ٔ as Tassy prints it (ḥulla'ī = حلهٔ).
- Flagged uncertain (shown to the reader): v. 708 nishāndan (Ganjoor fishāndan); v. 3783 yakhcha'ī (Ganjoor yā yakhī); v. 3928 maydān (or miyān). Re-checked on a zoomed image and confirmed: v. 667 murīd (Ganjoor barīd), v. 4207.
- To do: a human proofreading against the images; collation with the 1883/1891 lithographs and, by locator only, with Shafīʿī-Kadkanī (©).
3. Layers
| Layer | Source | Status |
|---|---|---|
| Persian | Tassy 1857, our transcription | public domain base; ours |
| English, per couplet | AI (Claude), scripts/persian/ai/mantiq-tayr-en.tsv, written from Tassy's Persian; no copyrighted translation consulted | AI, marked |
| Word glosses | AI (Claude), scripts/persian/ai/mantiq-tayr-glosses.tsv: 1,122 written forms, one gloss each, in the poem's sense | AI, marked |
| Transliteration, per hemistich and per word | AI (Claude), scripts/persian/ai/mantiq-tayr-tr.tsv: a reading in classical Persian as the metre scans it (§3a); checked by scripts/persian/scan_metre.py | AI, marked; 549/564 hemistichs scan, the rest marked † |
| FitzGerald, Bird-Parliament | 1889 (written 1857–62); archive.org BirdParliamentEnglishFareeduddinAttaarFitzgerald (the sacred-texts transcription); line numbers and page numbers removed, OCR slips fixed in build_mantiq_tayr.py | public domain; a free paraphrase, shown per section as a contrast |
| Masani, The Conference of the Birds (1924) | archive.org in.ernet.dli.2015.128685 | PD only in the US (Masani d. 1966: © in life+70 countries to 2036). Not shown. |
| Garcin de Tassy, French (1863) | archive.org manticuttaroul00arfauoft | public domain, verse-numbered like the Persian: the natural second translation for triangulation (not yet added) |
Build: python3 scripts/persian/build_mantiq_tayr.py → app/src/data/mantiq-tayr.json.
3a. Transliteration: scheme, method, and the lines that do not scan
Tassy's Persian has no short vowels, so any transliteration is a reading. The one shown (Aa → Transliteration, on by default; under each hemistich, and each word's form in the word panel) gives classical Persian as the metre requires it.
Scheme: IJMES Persian (stated on the About page, #transliteration):
- vowels a i u / ā ī ū; diphthongs ay aw (classical, not modern ey ow); majhūl ē ō not marked (written ī ū);
- Persian letters p ch zh g; و consonant v (w only in aw and in Arabic quotations); silent vāv of خو = khv (khvīsh, khvud);
- Arabic letters th dh ḥ kh ṣ ḍ ṭ ẓ ʿ gh q, hamza ʾ (not word-initial). Deviation from IJMES Persian, deliberate: ث ذ ض are th dh ḍ (as in the rest of the reader) rather than s̱ ẕ ż; ذ in Persian words is written dh as the letter (gudhashtan, not modern guzashtan);
- iḍāfa -i / -yi (after a vowel or silent h); indefinite -ī, after silent h -ʾī (ḥulla-ʾī);
- hyphens join prefixes (mī-, bi-, na-, ma-), enclitics and contractions (z-ān, k-ū, v-ar); ba- the preposition, bi- the verbal prefix; و after a consonant = u; Arabic phrases as said in the line (bismi llāh, Kalīmu llāh, hal min mazīd);
- one transliterated word for each Persian word as Tassy spaces it (so the word panel can show the reading in its place); capitals for proper names only.
Method. Written by AI (Claude, 3 October 2026) in four batches (vv. 658–724; 3202–3334; 3456–3936; 4104–4229), reading each hemistich against ramal musaddas maḥdhūf so that the metre fixes vowel length, the iḍāfa and the enclitic vowels (e.g. v. 664 hudhud-i āshufta dil, the iḍāfa demanded by the metre; v. 668 ṣāḥib asrār without one; v. 712 bigdhasht; v. 3226 na-hrāsad; v. 3319 murtadī with a single d; vv. 3574, 3578 bi-rikht with short i). Then validated mechanically by scripts/persian/scan_metre.py, which syllabifies the Latin and fits it to – u – – | – u – – | – u – (faʿilātun u u – – allowed in the first two feet, faʿilun u u – in the last), with these licences only: the last syllable of a hemistich counts long; CVVC/CVCC count long + short, except long vowel + n; a short word-final vowel may count long; a long word-final vowel (or one before the enclitic -yi/-ʾi) may count short before a vowel; a vowel-initial word may or may not take the consonant before it; a word-final double consonant may be said once or held. Negative controls (a hemistich with a word dropped or the iḍāfa removed) fail. Nothing was forced: where no honest reading scanned, the line was left and is listed.
Result: 282/282 couplets, 564 hemistichs; 549 scan, 15 do not (13 couplets). The reader marks them †.
| Verse | Hemistich (as read) | Note |
|---|---|---|
| 708.1 | mardī bāyad tamām īn rāh rā | one syllable short. Ganjoor's mard mī-bāyad would scan: re-check Tassy's image (this verse is already flagged for nishāndan) |
| 715.1 | ān parr aknūn dar nigāristān-i Chīn-ast | one syllable over as read |
| 715.2 | uṭlubu l-ʿilma wa-law bi-ṣ-Ṣīn az-īn-ast | the hadith line; does not fit as Arabic is said |
| 3217.1, .2 | dar miyān-i khūnat … / v-az hama bīrūnat bāyad āmadan | both halves fail at the same place (-ūnat bāyad); the reading is not found |
| 3479.1 | gar na-dārī shādī az vaṣl-i yār | one syllable short |
| 3567.1 | ṣad hazārān ṭifl sar burīda gasht | does not fit (Ganjoor sar bi-burīda does not either) |
| 3577.1 | gar na-mānad az dīv u az mardum athar | one syllable over; a v-az (Ganjoor vaz) for u az would remove it |
| 3577.2 | az sar-i yak qaṭra-yi bārān dar-gudhar | one syllable over |
| 4138.2 | na tanashān mānda u na par mānda-ʾī | one syllable over |
| 4147.1 | guft ān chāvush k-ay sargashtagān | the second foot does not fit |
| 4153.1 | z-ū kasī rā khvārī hargiz buvad | one syllable short |
| 4214.2 | bī tafakkur dar tafakkur māndand | short at the end; the rhyme with āmadand suggests a reading not found here |
| 4217.2 | k-āyina ast īn ḥaḍrat chūn āftāb | does not fit as read |
| 4220.1 | gar chihil u panjāh murgh āyand bāz | does not fit as read |
These are the first places to look in the human proofreading: several (708, 3577) point to a word that may be misread or that Tassy prints differently from the metre.
4. Notes on the text
- v. 715 quotes the hadith uṭlubū l-ʿilm wa-law bi-l-Ṣīn (Arabic, in the Persian line).
- v. 3476 quotes Q 50:30 hal min mazīd ("Is there more?"). The hoopoe's speech (vv. 667–679) alludes to Q 27:20–28 (Solomon, the hoopoe's absence, the letter to Sheba). Not yet run through
scripts/quran_refs.py, which reads Arabic texts only (§15: to adapt). - v. 4228 mā bi-sī murghī bisī awlā-tarīm: the pun the whole poem rests on, sī murgh "thirty birds" / Sīmurgh; the AI English gives the reading "than thirty birds" and says so.
- Metre: ramal musaddas maḥdhūf (fāʿilātun fāʿilātun fāʿilun) [REPORTED], one metre for the whole poem. Scanned through the transliteration (§3a): 549/564 hemistichs fit.
5. Resonances (for §16, not yet run)
Ibn Sīnā's Risālat al-Ṭayr (birds, journey, King: shared motif, documented reading of Ibn Sīnā by ʿAṭṭār not established here: parallel); Suhrawardī's Ghurba (the hoopoe's letter); al-Ghazālī's polished mirror and Ibn ʿArabī's mirror (heart-sight §3.3–3.4); the Poimandres' Anthropos and his reflection (contrast: there the reflection draws the Man down; here it is the goal).
6. Open questions
- Human proofreading of the transcription; the three flagged readings.
- Triangulation (§17) once Tassy's French (1863) is added; at present the only published English is FitzGerald's paraphrase, which is not couplet-aligned, so
scripts/triangulate.pyhas nothing to compare. - Persian lemmatisation (Hazm) and Steingass for the word panel. The metre scanner exists (§3a); a human check of the transliteration (especially the 15 hemistichs that do not scan) is still to do.
7. Change log
- 2026-10-03: licence resolved (Ganjoor unclear → not shown; Tassy 1857 public domain → base); 282 couplets transcribed; AI English and glosses; FitzGerald per section; reader
/read/mantiq-tayr. - 2026-10-03: transliteration layer (§3a): IJMES Persian, AI, all 282 couplets, per hemistich and per word;
scan_metre.pyvalidates it against the metre (549/564 scan; 15 listed, not forced); Aa → Transliteration toggle, on by default.